JP3563553B2 - Genetic algorithm analysis display processing device - Google Patents

Description

【００８３】
表１を簡単に説明する。Ｎｏ１の染色体は変数１が「７５」、変数２が「２０４」で構成され、その適応度は「９９３」であることを示す。そして「７５」と「２０４」をそれぞれ２進数で示したものが染色体として表示された「０１００１０１１」、「１１００１１００」である。
【００８４】
Ａ．表示図の説明
（１）染色体集団のクラスタ樹状図
本発明の説明で用いる表示図である図３〜図１２を説明する。
【００８５】
図３は、表１に示す染色体集団のクラスタ樹状図である。一般に、遺伝アルゴリズムの染色体の間では「染色体が似ている」、「染色体が似ていない」を数値に表す類似度と呼ぶ量を定義できる。更に、ある類似度数値以下の染色体の集まりをクラスタと呼ぶ。１個の染色体も特別なクラスタとみなせる。一般的にクラスタの間でも同様な類似度が定義できる。
【００８６】
クラスタ樹状図は、図３に示す如く、染色体間の類似度を横軸にし、縦軸に染色体を並べる。複数個のクラスタもしくは染色体ｃｌ_１ 、ｃｌ_２ 、ｃｌ_３ ・・・の間の類似度がＤであるとき、縦軸上のクラスタからの横線を、横軸の類似度値がＤの所で縦線で結ぶ。図３では染色体６と７の類似度が４、染色体８と１０の類似度が５等の例を示す。クラスタ樹状図は、染色体集団内でのバラツキ具合を知るのに有効である。染色体集団のクラスタ樹状図の作成手法は、本発明者等による前記平成７年特許願第２３８３８９号に説明ずみである。また本発明においては表示図上で染色体あるいは染色体のクラスタや世代がマウスの如き位置指定装置で指示できればよい。
【００８７】
（２）染色体集団の変数定義域分布図
図４に示す如く、染色体１、２・・・を例えば２つの変数で表示するものである。例えば表１に示す如く、染色体１、２、３・・・を１６ビットで表現するとき、初めの８ビットの示す１０進数で変数１を示し、後の８ビットの示す１０進数で変数２を示し、この変数１と変数２の組み合わせで図４に示す如き変数定義域分布図を作るものである。一般的には、染色体はｎ変数で表示され、ｎ変数の中から適当な２変数を選び、それら２変数の張る２次元の定義域をＸ、Ｙ軸として各染色体の表す変数値に対応する位置にマーク図形で表示するものである。
【００８８】
例えば表１の染色体集団を図４の如く表示したとき、これにより染色体２、３、７、９を囲む領域付近に何かピークを示す変数の組み合わせの存在が示唆され、染色体１、４、８、９を囲む領域付近に何かピークを示す変数の組み合わせの存在が示唆され、染色体５、６を囲む領域付近に同様にピークを示す変数の組み合わせの存在が示唆される。
【００８９】
（３）染色体集団適応度分布の世代変化図
図５に示す如く、横軸に世代、縦軸に適応度をとり、各世代での集団内の染色体全部を点表示したものであり、各点が１つの染色体を表す。
【００９０】
図５では、０世代における１０個の染色体が１世代、２世代・・・と経る毎に染色体の適応度がどのように変化するのかを示している。遺伝アルゴリズムの演算結果により染色体の適応度は変化し、等しい場合もある。１世代では染色体の適応度の一部に重なりが存在し、１４世代以降は適応度が全部重なるものとなる。前記特願平７−２３８３８９号に示す如く、隣合う世代間で染色体を線で結ぶこともあるが、結線の有無は本発明にとって重要でないので、図５では省略している。
【００９１】
（４）変数１値の分布の世代変化図
図６に示す如く、横軸に世代、縦軸に染色体の表す変数１（例えば表１の染色体の前半の８ビットの１０進数）の値をとり、各世代での集団内の染色体の変数１の値を点表示したものである。これにより世代を経るとともに集団内で変数１の値の分布の世代変化、つまり収束状態を示すことができる。
【００９２】
（５）最大適応度の世代変化図
図７に示す如く、横軸に世代、縦軸に染色体の適応度をとり、各世代での集団内の染色体の適応度の最大値を点表示したものである。例えば図５に示す如く、各世代での集団内の染色体の適応度が示されるとき、各世代での適応度の最大値を図７では表示する。これにより世代を経るとともに、どの程度最適な解を見つけているか、世代変化の様子を示すことができる。
【００９３】
（６）適応度最大値と適応度偏差の世代変化軌跡図
図８に示す如く、横軸に染色体集団内での適応度最大値をとり、縦軸を適応度標準偏差値として、各世代での値を点表示し、世代順に線で結んだものである。これにより世代を経るとともに適応度の分布状況を軌跡の形で世代変化の様子を示すことができる。
【００９４】
（７）異なる実行事例での適応度値の分布図
図９に示す如く、横軸に集団内の最大適応度値をとり、縦軸に適応度標準偏差値をとり、いくつかの異なる条件下で実行した実行事例について、ある特定世代での適応度の最大値と偏差値を点表示したものである。図９は疑似乱数の初期値が異なる６つの実行事例について示している。これにより条件の違いによる結果の違いを示すことができる。
【００９５】
（８）異なる実行事例での変数定義域値分布図
図１０に示す如く、ある２変数の張る２次元の定義域をＸ、Ｙ軸として、異なる条件下での実行事例について、ある特定世代についての、集団内での最良染色体つまり集団内で適応度値が最大の染色体の変数値を点表示したものである。これにより例えば疑似乱数の初期値の違いによる結果の違いを示すことができる。
【００９６】
（９）最良染色体の世代変化図
図１１に示す如く、横軸に世代、縦軸に染色体の適応度値をとり、いつくかの遺伝アルゴリズムの実行事例について各世代での最良染色体つまり集団内で最大の適応度を持つ染色体を点表示し、結線したものである。各点が１染色体を示す。第１１世代で実行事例１、２の最大の適応度が重なり、第１４世代で実行事例１、３の最大の適応度が重なる。これにより異なる条件下で計算した実行事例間での相違を知ることができる。
【００９７】
（１０）適応度偏差の世代変化図
図１２に示す如く、横軸に世代、縦軸に染色体集団全体の適応度値の標準偏差をとり、いくつかの遺伝アルゴリズムの実行事例について、各世代での集団全体の適応度値の標準偏差を点表示し、結線したものである。第１６世代で実行事例１、２の標準偏差値が重なる。これにより世代を経るとともに、適応度値の分散つまり集団のばらつき具合の世代変化の様子を示すことができる。
【００９８】
Ｂ、本発明の動作説明
（１）本発明の動作説明（その１）
図２において、補助記憶装置３に、例えば前記図１５により説明した遺伝アルゴリズムによる演算結果を格納しておく。そして文字入力装置４から表示コマンドを入力すると第１表示部６−１では、この補助記憶装置３に格納された演算結果により、その表示図作成手段９−１が図３に示す如きクラスタ樹状図を作成し、表示図制御手段１２−１を経由して制御部５に送る。
【００９９】
制御部５はこのクラスタ樹状図を表示画面１−０の定められた位置に表示する画像表示装置１に伝達する。画像表示装置１は、これを表示画面１−０のあらかじめ第１表示部６−１用に定められたサイズに縮小し、これに表示画面１−０の座標と図番を付加して領域に、図番例えばＧ１を付記して表示する。なお、前記表示図作成手段９−１により作成されたクラスタ樹状図は、別に記憶手段１３−１に格納される。
【０１００】
また第２表示部６−２では、補助記憶装置３に格納された演算結果により、その表示図作成手段９−２が、図４に示す如き染色体集団の変数定義域分布図を作成し、同様にして、画像表示装置１により表示画面１−０のあらかじめ定められた領域に、図番例えばＧ２を付記して表示する。またこの染色体集団の変数定義域分布図は別に記憶手段１３−２に格納される。
【０１０１】
このようにして各表示部が動作し、表示画面１−０に各図が表示される。
次にオペレータが染色体集団のクラスタ樹状態と染色体集団の変数定義域分布状態との関係を観察したいと判断したとき、文字入力装置４から表示画面選択信号ＳＧ１、ＳＧ２を入力する。これにより制御部５は第１表示部６−１の表示図制御手段１２−１と第２表示部６−２の表示図制御手段１２−２にこれを伝達して選択表示動作状態にする。もし、文字入力装置４から表示画面選択信号の入力がなければ、全表示図が選択対象となる。
【０１０２】
そしてオペレータが表示画面１−０上の図番Ｇ１のクラスタ樹状図の染色体１０（図３参照）のマーク図形をマウスの如き位置指定装置２により指示すると、その位置が画像表示装置１により検知され、制御部５に伝達される。制御部５は、その位置を前記第１表示部６−１の表示図制御手段１２−１から伝達された、元の図３に示すクラスタ樹状図の座標位置つまり元座標位置信号に変換し、この元座標位置信号を第１表示部６−１の表示図制御手段１２−１に伝達する。なおこの元座標位置信号への変換は制御部５ではなく表示図側で行うことも可能である。
【０１０３】
第１表示部６−１の表示図制御手段１２−１では、この元座標位置信号により記憶手段１３−１をアクセスしてこれが染色体１０であることを認識する。そして表示図指示物送信手段７−１に対して、染色体１０を区別表示することを他の表示部に対して通知するように指示する。これにより表示図指示物送信手段７−１は、他の表示部である第２表示部６−２〜第ｎ表示部６−ｎに対して染色体１０を区別表示することを指示する送信信号を制御部５を経由して送出する。なお前記認識は表示図指示物送信手段７−１で行うこともできる。
【０１０４】
このとき動作状態にあるのは第２表示部６−２だけであるのでこの送信信号はその表示図制御手段１２−２を経由して表示図指示物受信手段８−２により受信され、受信信号が解読されて表示図制御手段１２−２は表示図指示物図示手段１１−２に対して染色体１０を区別表示すべきことを指示する。
【０１０５】
これにより、表示図指示物図示手段１１−２は記憶手段１３−２をアクセスして、その図４に示す染色体集団の変数定義域分布図における染色体１０の位置を読み出し、これを区別表示すべきことを表示図制御手段１２−２を経由して制御部５に伝達する。また表示図指示物図示手段１１−２が、表示図作成手段９−２と協力して区別表示された表示図全体を作り、それを制御部５を介して画像表示装置に表示させることもできる。
【０１０６】
これにより制御部５は、図番Ｇ２における染色体１０の位置が表示画面１−０のどの位置になるのかを演算して、その演算した区別表示すべき位置を画像表示装置１に伝達し、図番Ｇ２における染色体１０のマーク図形が例えばブリンクされるとか、別の色でカラー表示されるとか、模様が付けられるとか、矢印図形が付加される等の区別表示される。
【０１０７】
このようにして、オペレータは図番Ｇ１の染色体に対応する図番Ｇ２上の染色体の表示状態を観察することができ、クラスタ樹状態と変数定義域分布状態との対応関係が観察できる。
【０１０８】
同様にしてクラスタ樹状図上の染色体１、８、１０からなるクラスタを指示する横線を位置指定装置２により指示すると、染色体集団の変数定義分布図上の染色体番号１、８、１０のマーク図形がすべて区別表示される。
【０１０９】
逆に、図番Ｇ２上の染色体集団の変数定義域分布図上の、例えば染色体４のマーク図形を位置指定装置２で指示すると、同様にして図番Ｇ１上のクラスタ樹状図上の染色体４のマーク図形が区別表示される。
【０１１０】
また図番Ｇ２上の染色体集団の変数定義域分布図上の、例えば染色体番号６、７のマーク図形を円形等の図形でグループ指示することにより、図番Ｇ１上のクラスタ樹状図上の染色体番号６、７のマーク図形が区別表示される。
【０１１１】
なお、一般に、クラスタ樹状図や染色体集団の変数定義域分布図において、染色体のそれぞれに番号が振られているとは限らない。図３及び図４では説明の都合で番号を記述しているものである。
【０１１２】
（２）本発明の動作説明（その２）
表示画面１−０に表示された表示図から染色体集団適応度分布の世代変化図（図５）と染色体の変数１の値の分布の世代変化図（図６）を使用する場合である。
【０１１３】
例えば第８世代の適応度分布の世代変化図上のある染色体、例えば染色体５（図５では番号省略）を位置指定装置２により指示すると、前記と同様にして、第３表示部（図示省略）の表示図指示物送信手段から第８世代の染色体５を区別表示することを指示する送信信号が出力される。
【０１１４】
これが図６の表示図を作成する第４表示部（図示省略）の表示図指示物受信手段で受信され、表示画面１−０上の染色体の変数１の値の分布の世代変化図の第８世代の染色体５の変数１の値が区別表示、例えばブリンク表示される。
【０１１５】
逆に、表示画面１−０上の変数１の値の分布の世代変化図の第８世代の染色体５を、位置指定装置により指示すると、染色体集団適応度分布の世代変化図上の第８世代の染色体５に対応するマーク図形が区別表示される。
【０１１６】
このようにして定義域上の変数１の値の分布と、値域上の適応度の分布との対応関係を知ることができる。
（３）本発明の動作説明（その３）
表示画面１−０に表示された表示図から染色体集団のクラスタ樹状図（図３）と染色体集団適応度分布の世代変化図（図５）を使用する場合である。例えば第８世代の染色体集団のクラスタ樹状図上の染色体５を位置指定装置２で指示すると、前記と同様にして、第１表示部６−１の表示図指示物送信手段７−１から第８世代の染色体５を区別表示することを指示する送信信号が出力される。
【０１１７】
これが図５の表示図を作成する第３表示部（図示省略）の表示図指示物受信手段で受信され、表示画面１−０上の染色体集団適応度分布の世代変化図の第８世代の染色体５に対応するマーク図形が区別表示される。
【０１１８】
また染色体集団のクラスタ樹状図上の染色体６、７から成るクラスタを指す横線を位置指定装置２で指示すると、染色体集団適応度分布の世代変化図の染色体６、７に対応するマーク図形がすべて区別表示される。
【０１１９】
逆に表示画面１−０上の染色体集団適応度分布の世代変化図の例えば第６世代の染色体４のマーク図形を、位置指定装置２で指示すると、第６世代の染色体集団のクラスタ樹状図上の染色体４のマーク図形が区別表示される。このとき、表示画面１−０に表示中の染色体集団のクラスタ樹状図が別の世代のもの、例えば第８世代のものであるとき、次のような制御が行われる。
【０１２０】
すなわち、第１表示部６−１の表示図指示物受信手段８−１が第６世代の染色体４を区別表示することを指示した信号を受信すると、これが解読される。こととき表示図制御手段１２−１は表示画面１−０に表示中のものが第８世代のものであり、第６世代のものでないことを認識するので、制御部５に対して補助記憶装置３から第６世代の演算結果読み出し、伝送を求める。
【０１２１】
そしてこれにより表示図作成手段９−１が第６世代の染色体集団のクラスタ樹状図を作成し、制御部５を経由して画像表示装置１に送出し、今度は表示画面１−０に第６世代の染色体集団のクラスタ樹状図が表示される。このとき記憶手段１３−１にこれが格納される。
【０１２２】
続いて第６世代のこのクラスタ樹状図の該当する染色体４の位置を表示図作成手段９−１が表示図制御手段１２−１に指示するので、表示図指示物図示手段１１−１はこれを区別表示すべきことを制御部５に指示する。制御部５では染色体４の位置を演算してこれを区別表示するよう画像表示装置１に指示する。このようにして表示中のものと別世代のものが指示されても、指示されたものを区別表示できる。
【０１２３】
このようにしてクラスタ樹状図と染色体集団適応度分布の世代変化図との対応関係を知ることができる。
（４）本発明の動作説明（その４）
表示画面１−０に表示された表示図から、最大適応度の世代変化図（図７）と適応度最大値と適応度偏差の世代変化軌跡図（図８）を使用する場合である。例えば最大適応度の世代変化図の第３世代のマーク図形を位置指定装置２で指示すると、前記と同様にして第５表示部（図示省略）の表示図指示物送信手段から第３世代を区別表示することを指示する送信信号が出力され、これが図８の表示図を作成する第６表示部（図示省略）の表示図指示物受信手段で受信され、表示画面１−０上の適応度最大値と適応度偏差の世代変化軌跡図の第３世代に対応するマーク図形が区別表示される。
【０１２４】
逆に、適応度最大値と適応度偏差の世代変化軌跡図上の第６世代のマーク図形を位置指定装置２で指示すると、最大適応度の世代変化図の第６世代のマーク図形が区別表示される。
【０１２５】
これにより最大適応度の世代変化状態と、適応度の世代変化軌跡状態を観察することができる。
（５）本発明の動作説明（その５）
表示画面１−０に表示された表示図から染色体集団適応度分布の世代変化図（図５）と適応度最大値と適応度偏差の世代変化軌跡図（図８）を使用する場合である。例えば染色体集団適応度分布の世代変化図の座標目盛上で、第７世代を位置指定装置２で指示すると、前記と同様にして第３表示部（図示省略）の表示図指示物送信手段から第７世代を区別表示することを指示する送信信号が出力され、これが図８の表示図を作成する第６表示部（図示省略）の表示図指示物受信手段で受信され、表示画面１−０上の適応度最大値と適応度偏差の世代変化軌跡図の第７世代に対応するマーク図形が区別表示される。
【０１２６】
逆に適応度最大値と適応度偏差の世代変化軌跡図上の第４世代のマーク図形を、位置指定装置２で指示すると、前記と同様にして第６表示部（図示省略）の表示図指示物送信手段から第４世代を区別表示することを指示する送信信号が出力され、これが図５の表示図を作成する第３表示部（図示省略）の表示図指示物受信手段で受信され、これが解読されて表示図指示物図示手段にこの第４世代を区別表示することが伝達される。
【０１２７】
これにより表示図指示物受信手段は記憶装置をアクセスして、第４世代のすべての染色体の位置を読み出し、これらを区別表示することを表示制御手段を経由して制御部５に伝達する。
【０１２８】
これにもとづき制御部５は、第４世代のすべての染色体の位置を表示画面１−０上の位置に演算してこれらを画像表示装置１に伝達する。このようにして染色体集団適応度分布の世代変化図における第４世代に属するすべての染色体のマーク図形が区別表示される。
【０１２９】
このようにして染色体集団適応度分布の世代変化と、適応度最大値と適応度偏差の世代変化軌跡状態を観察することができる。
（６）本発明の動作説明（その６）
表示画面１−０に表示された表示図から異なる実行事例での適応度値の分布図（図９）と異なる実行事例での変数定義域値分布図（図１０）を使用する場合である。例えば特定世代の、異なる実行事例での適応度値の分布図上で実行事例２のマーク図形を位置指定装置２で指示すると、前記と同様にして第７表示部（図示省略）の表示図指示物送信手段から実行事例２を区別表示することを指示する送信信号が出力される。
【０１３０】
これが図１０の表示図を作成する第８表示部（図示省略）の表示図指示物受信手段で受信され、表示画面１−０上の、同じ特定世代の、異なる実行事例での変数定義域値分布図の実行事例２に対応するマーク図形が区別表示される。
【０１３１】
逆に特定世代の異なる実行事例での変数定義域値分布図（図１０）上の実行事例３のマーク図形を位置指定装置２で指示すると、同じ特定世代の、異なる実行事例での適応度値の分布図（図９）上の実行事例３に対応するマーク図形が区別表示される。
【０１３２】
これにより異なる実行事例での適応度値の分布状態と、異なる実行事例での変数定義域値分布状態を観察することができる。
（７）本発明の動作説明（その７）
表示画面１−０に表示された表示図から最良染色体の世代変化図（図１１）と適応度偏差の世代変化図（図１２）を使用する場合である。例えば最良染色体の世代変化図上で矩形の点線枠内で表示した但し書き部分の白丸部分あるいは「実行事例１」という文字部分を位置指定装置２で指示すると、前記と同様にして第９表示部（図示省略）の表示図指示物送信手段から実行事例１を区別表示することを指示する送信信号が出力される。
【０１３３】
これが図１２の表示図を作成する第１０表示部（図示省略）の表示図指示物受信手段で受信され、表示画面１−０上の、適応度偏差の世代変化図の実行事例１に対応するマーク図形及びその接続線が区別表示される。
【０１３４】
逆に適応度偏差の世代変化図（図１２）上の矩形の点線枠内でも表示された但し書き部分の黒丸部分あるいは「実行事例２」という文字部分を位置指定装置２で指示すると、最大適応度世代変化図（図１１）上の実行事例２に対応するマーク図形及びその接続線が区別表示される。
【０１３５】
なお位置指定装置２による指示は前記但し書き部分内でなくとも接続線を指示してもよい。
これにより同一実行事例における最大適応度世代変化状態と適応度偏差の世代変化状態とを観察することができる。
【０１３６】
（８）本発明の動作説明（その８）
表示画面１−０に表示された表示図から、異なる実行事例での変数定義域値分布図（図１０）と、最良染色体の世代変化図（図１１）を使用する場合である。例えば第４世代での異なる実行事例での変数定義域値分布図上で実行事例１のマーク図形を位置指定装置２で指示すると、前記と同様にして第８表示部（図示省略）の表示図指示物送信手段から第４世代の実行事例１を区別表示することを指示する送信信号が出力される。
【０１３７】
これが図１１の表示図を作成する第９表示部（図示省略）の表示図指示物受信手段で受信され、表示画面１−０上の最良染色体の世代変化図の実行事例１の第４世代に対応するマーク図形が区別表示される。
【０１３８】
逆に最良染色体の世代変化図（図１１）上の第５世代目の実行事例３のマーク図形を位置指定装置２で指示すると、第５世代での異なる実行事例での変数定義域値分布図上の実行事例３に対応するマーク図形が区別表示される。表示図上の異なる実行事例での変数定義域値分布図が第５世代でないとき、前記動作説明（その３）と同様に、第５世代の異なる実行事例での変数定義域値分布図が表示されたのち区別表示する。
【０１３９】
これにより特定世代における異なる実行事例での変数定義域値分布状態と、その実行事例における最良染色体の状態とを観察することができる。
（９）本発明の動作説明（その９）
表示画面１−０に表示された表示図から、最良染色体の世代変化図（図１１）と適応度偏差の世代変化図（図１２）を使用する場合である。例えば最良染色体の世代変化図上で世代座標目盛上の、第２世代を位置指定装置２で指示すると、前記と同様にして第９表示部（図示省略）の表示図指示物送信手段から各実行事例の第２世代を区別表示することを指示する送信信号が出力される。
【０１４０】
これが図１２の表示図を作成する第１０表示部（図示省略）の表示図指示物受信手段で受信され、表示画面１−０上の適応度偏差の世代変化図の、第２世代に属するすべての実行事例即ち実行事例１、２、３の各第２世代のマーク図形が区別表示される。
【０１４１】
逆に適応度偏差の世代変化図上の世代座標目盛上の例えば第３世代を位置指定装置２で指示すると、前記と同様にして、最良染色体の世代変化図上の、第３世代に属するすべての実行事例のマーク図形が区分表示される。
【０１４２】
これにより最良染色体の世代変化と、適応度偏差の世代変化の状態を観察することができる。
Ｃ、本発明の他の実施の形態
前記説明は、いずれもビット列表現型染色体の例について説明したが、本発明は勿論これに限定されるものではなく、順序表現型染色体についても適用できるものである。図１３により順序表現型染色体表現の場合の染色体分布図について説明する。
【０１４３】
図１３は順序表現型染色体表現の場合の１つの例として都市Ａ〜Ｈに対する巡回セールスマン問題に対するものであり、定義域上での染色体の分布を図示したものであり、特定世代における各染色体が図１３（Ａ）に示す如く、染色体１＝「ＥＢＡＤＣＧＨＦ」で、染色体２＝「ＧＥＦＢＣＡＤＨ」で、・・・染色体８＝「ＧＤＣＡＥＢＦＨ」でそれぞれ表現されている。
【０１４４】
図１３（Ｂ）は染色体１〜８の２都市間の隣接性を示すものであり、隣接都市を黒マークで、非隣接都市を白マークで示しており、例えば染色体１では巡回路上で都市Ａと都市Ｂは隣り合っているので黒マークしており、都市Ａと都市Ｃは隣り合っていないので白マークしている。同様に染色体２では巡回路上で都市Ａと都市Ｂは隣り合っていないので白マークしており、都市Ａと都市Ｃは隣り合っているので黒マークしている。したがって各染色体１〜８において黒マークの多い都市ＡＢ、ＡＤ、ＢＥ、ＧＨ等は、隣接性の高いものであることがわかる。
【０１４５】
このような順序表現型染色体を用いて、図３に示す如き染色体集団のクラスタ樹状図を作成する手法については、前記特願平７−２３８３８９号に記載されている。
【０１４６】
いま、表示画面１−０に図１３（Ｂ）と、これに基づき作成したクラスタ樹状図を表示するとき、図１３（Ｂ）の染色体番号を示す横軸目盛、もしくは図１３（Ｂ）中のマーク部分を位置指定装置２で指示すると、横軸目盛の場合には目盛の番号の染色体を、マークの場合にはこのマークを含む染色体の番号を区別表示することを、この順序表現型染色体の２都市間の隣接性を示す図を表示している表示部の表示図指示物送信手段からクラスタ樹状図側の表示部の表示図指示物受信手段に送り、同じ世代のクラスタ樹状図の対応する染色体を区別表示する。
【０１４７】
逆にクラスタ樹状図上の染色体が指示すると、前記と同様にして２都市間の隣接性を示す図の対応する染色体のマーク全部、もしくは横軸目盛の染色体番号が区別表示される。
【０１４８】
これにより染色体集団のクラスタ樹状図と、順序表現型染色体の２都市間の隣接性の状態とを観察することができる。
本発明では順序表現型染色体表現として、また図１４において横軸を実行事例１、２、３・・・の最良染色体とし、これによりその異なる条件下での実行事例においてある特定世代についての集団内での適応度が最大の染色体の隣接性を示すことができる。これを前記図１０の代わりに使用することもできる。
【０１４９】
なお前記説明では一方の図面を指示することにより他方の１つの図面について区別表示される例について説明したが、本発明では勿論これに限定されるものではなく、１つの図面における指示により複数の図面において対応部分を区別表示することも可能である。
【０１５０】
前記説明では、補助記憶装置３に各世代毎の染色体集団やこれにより作成された各種図形が格納された例について説明したが本発明は勿論これに限定されるものではなく、主記憶装置にこれらのものを格納することもできる。
【０１５１】
また図示省略した遺伝アルゴリズム演算装置により前記図１５、図１６に示す如き遺伝アルゴリズム演算処理を行って得られたデータを使用する例について説明したが、本発明は勿論これに限定されるものではなく、制御部５により遺伝アルゴリズム演算処理を行って得られるデータを使用することもできる。
【０１５２】
本発明によれば遺伝アルゴリズムの実行結果を、複数の図面を対比しながらその進行状態を観察できるので、実行結果を予測するための判断が正確になる。
【０１５３】
【発明の効果】
本発明によれば下記の如きすぐれた効果が得られる。
（１）表示画面に複数種類の表示図を表示しておき、ある表示図上の図形の１部を位置指定手段により指示したとき、他の表示図上の対応図形を、色、模様、点滅、矢印図形などの視覚的手法等の区別表示により表示したので、例えば染色体集団のクラスタ樹状図と染色体集団の変数定義域分布図のような、複数種類の表示図上での同一染色体等の染色体関連情報を観察することができるので、染色体情報の特徴、例えば染色体集団分布の特徴等を理解するのを助けることができる。
【０１５４】
（２）表示画面にある範囲の世代の染色体集団の適応度分布の世代変化図と染色体の別の種類の分布の世代変化図を表示しておき、その対応関係を観察することができるので、複数種類の表示図上での同一染色体の位置関係を容易に認識することができ、それにより染色体集団分布の特徴を理解するのを助けることができる。
【０１５５】
（３）表示画面に特定世代の染色体集団のクラスタ樹状図のような特定世代についての表示図と染色体集団適応度分布の世代変化図のような世代範囲についての表示図を表示しておき、その対応関係を観察することができるので、複数種類の表示図上での同一染色体の位置関係を容易に認識でき、それにより染色体集団分布の特徴を理解するのを助けることができる。
【０１５６】
（４）表示画面に、最大適応度の世代変化図と、適応度最大値と適応度偏差の世代変化軌跡図のような遺伝アルゴリズムの染色体集団の各世代における解析量をある世代範囲について複数種類表示しておき、その対応関係を観察することができるので複数種類の表示図上での、同一世代の解析量の位置関係を容易に認識でき、それにより染色体集団の解析量の世代変化の特徴を理解するのを助けることができる。
【０１５７】
（５）表示画面に、染色体集団適応度分布の世代変化図と適応度最大値と適応度偏差の世代変化軌跡図のような染色体についてのある特定世代あるいは世代範囲での複数種類の表示図を表示しておきその対応関係を観察することができるので、複数種類の表示図上での同一世代の解析量の位置関係を容易に認識でき、それにより染色体集団の解析量の世代変化の特徴を理解するのを助けることができる。
【０１５８】
（６）表示画面に、異なる実行事例での適応度値の分布図と異なる実行事例での変数定義域値分布図のように、異なる条件下で実行した、遺伝アルゴリズムの複数の実行事例を表示しておきその対応関係を観察することができるので、複数種類の表示図上での同一実行事例の解析量の位置関係を容易に認識でき、それにより異なる条件下での実行事例の違いの特徴を理解するのを助けることができる。
【０１５９】
（７）表示画面に、最良染色体の世代変化図と適応度偏差の世代変化図のように、世代範囲についての実行事例の複数の種類の表示図を表示しておきその対応関係を観察することができるので、複数種類の表示図上での同一実行事例の解析量の位置関係を容易に認識でき、それにより異なる条件下での実行事例の世代変化の違いの特徴を理解するのを助けることができる。
【０１６０】
（８）表示画面に、異なる実行事例での変数定義域値分布図のようなある特定世代での表示図と、最良染色体の世代変化図のような世代範囲についての表示図を表示しておきその対応関係を観察することができるので、複数種類の表示図上での同一実行事例の解析量の位置関係を容易に認識でき、それにより異なる条件下での実行事例の違いを理解するのを助けることができる。
【０１６１】
（９）表示画面上に、最良染色体の世代変化図と適応度偏差の世代変化図のように、ある世代範囲についての実行事例の複数の種類の表示図を表示しておきその対応関係を観察することができるので、複数種類の表示図上で同一世代での実行事例の解析量の違いを容易に認識でき、それにより異なる条件下での実行事例の違いを理解するのを助けることができる。
【図面の簡単な説明】
【図１】本発明の原理構成図である。
【図２】本発明の一実施の形態である。
【図３】染色体集団のクラスタ樹状図である。
【図４】染色体集団の変数定義域分布図である。
【図５】染色体集団適応度分布の世代変化図である。
【図６】変数１の値の分布の世代変化図である。
【図７】最大適応度の世代変化図である。
【図８】適応度最大値と適応度偏差の世代変化軌跡図である。
【図９】異なる実行事例での適応度値の分布図である。
【図１０】異なる実行事例での変数定義域値分布図である。
【図１１】最良染色体の世代変化図である。
【図１２】適応度偏差の世代変化図である。
【図１３】順序表現型染色体表現の場合の染色体分布図である。
【図１４】異なる実行事例での最良染色体分布図である。
【図１５】順序表現の従来例説明図である。
【図１６】ビット列表現の従来例説明図である。
【符号の説明】
１ 画像表示装置
１−０ 表示画面
２ 位置指定装置
３ 補助記憶装置
４ 文字入力装置
５ 制御部
６−１ 第１表示部
６−２ 第２表示部
６−ｎ 第ｎ表示部
７−１ 送信部
７−２ 送信部
８−１ 受信部
８−２ 受信部
９−１ 表示図作成手段
９−２ 表示図作成手段
１０−１ 表示図指示物検出手段
１０−２ 表示図指示物検出手段
１１−１ 表示図指示物図示手段
１１−２ 表示図指示物図示手段
１２−１ 表示図制御部
１２−２ 表示図制御部
１３−１ 記憶手段
１３−２ 記憶手段[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a genetic algorithm analysis display processing device, in particular, in a plurality of types of display diagrams, the same chromosome, the same generation, the corresponding relationship of the same execution example, by visually clearly on the display diagram, the chromosome population It is intended to make it easy to understand the behavior and generational change of the game.
[0002]
[Prior art]
A genetic algorithm (hereinafter referred to as GA) converts a solution to the problem to be solved into a symbolic representation called chromosomes, and repeatedly performs an operation called a genetic operation on these chromosomes to obtain a group of chromosomes. Is a probabilistic calculation method for obtaining an optimal solution or a sub-optimal solution by gradually changing.
[0003]
For each of the two representative symbol string representations of the GA, FIGS. 15 and 16 show the stages of the standard genetic operation of the GA. FIG. 15 shows a genetic operation of GA using a symbol string expression method known as an ordered expression, and FIG. 16 shows a GA genetic operation using a bit string expression method known as a bit string expression.
[0004]
As shown in FIG. 15, a standard arithmetic device of this symbol string display method includes an initial chromosome group generation unit 100, a fitness evaluation operation unit 101, a selection operation unit 102, a cross operation unit 103, a mutation operation unit 104, and the like. It is configured. The operation of FIG. 15 will be described with respect to an example of generating a solution to the traveling salesman problem.
[0005]
The traveling salesman problem means that when there are n cities and a cost Pij between two cities Ci and Cj is given (i, j = 1, 2,... N), starting from a certain city, This is a problem of finding a tour that minimizes the total cost コ ス ト Pij (tour length) along a route in a tour that returns to the first city by traveling once through all the remaining cities. In the following description, an example in which n = 10 will be described. In addition, ten cities are indicated by numerals 0 to 9.
[0006]
The cost between cities 0 to 9 and two cities and necessary parameters are input to the initial chromosome population generation unit 100. The cost between the two cities is transmitted from the initial chromosome population generation unit 100 to the fitness evaluation calculation unit 101. Further, the parameters are selectively transmitted from the initial chromosome population generation unit 100 to the selection unit operation 102, the crossover operation unit 103, and the mutation operation unit 104 according to the parameters. The parameters include, for example, the probability of erasing low chromosomes while leaving one with high fitness, the position of crossover and its probability, the position of mutation, and its probability.
[0007]
The initial chromosome population generation unit 100 generates each chromosome a, b, c, d. For example, as shown in the group (1), chromosomes a = "2450976183", b = "3164250987", c = "90552816734", and d = "9162748350" are generated.
[0008]
These chromosomes were specifically generated as follows. Choose one city at random and call it C ₁ And Partial circuit C ₁ , C ₂ , ... C _k City C to visit next _{k + 1} And select the city with the lowest cost from among the uncirculated cities. This is sequentially repeated to form one complete tour. That is, it is created based on the conventional nearest neighbor method.
[0009]
As another generation method, a starting city is randomly selected, and a city with a lower cost is selected from uncirculated cities with a higher probability. This is repeated to make one tour.
[0010]
As still another generation method, a starting point city is selected at random, and a fixed-length partial tour connecting sequentially the cities having the lowest cost is created. With this subtour as the fixed part of the subpopulation, the remaining cities are selected at random and a tour is created.
[0011]
Of course, it is also possible to form a traveling route by sequentially selecting a city by a random method in which a certain city is randomly selected as a starting point, and one is also randomly selected from uncirculated cities. The method of generating chromosomes is not limited to these, and may be generated by other methods.
[0012]
The fitness evaluation calculation unit 101 calculates the fitness (a), (b), (c), (d) of each chromosome a, b, c, d. In the case of the traveling salesman problem, the one with the lowest total cost when passing through a city circuit indicated by chromosomes is the best chromosome.In this case, as the fitness, for example, the reciprocal of the total cost is calculated. Can be taken. At this time, the one with the smallest total cost is calculated as the one with the highest fitness.
[0013]
The selection operation unit 102 leaves chromosomes in the population of chromosomes a to d with high fitness and deletes chromosomes with low fitness. When this deletion is necessary, the one with high fitness is copied if necessary. At this time, the remaining erasure can be selected by determining the probability by, for example, selecting a roulette wheel.
[0014]
(2) in FIG. 15 indicates that when the fitness of the chromosomes a to d of the group (1) is calculated by the fitness evaluation operation unit 101, the fitness of the chromosome a is maximum and the fitness of the chromosome d is minimum. Therefore, the chromosome d is deleted by the selection operation unit 102, and the chromosome a is replicated as the chromosome d '.
[0015]
The crossover operation unit 103 crosses two chromosomes to create two chromosomes of different types. Specifically, a description will be given using a crossover method called PMX which can cross over the sequence of city names in the traveling salesman problem without contradiction. For example, using the chromosomes a and c of the group {circle around (2)} in FIG. 15, crossover is performed between the second to fifth from the top as shown in the explanation part 111 as shown by triangle marks.
[0016]
Therefore, first, the portions of the chromosomes a and c sandwiched between the triangular marks are exchanged. As a result, the chromosome a becomes “2052876183”. However, since two 2 and 8 exist, and four and nine do not exist, the first two existing in the chromosome a is replaced by four in order to make use of the crossed part. Then, 8 is converted to 9 to obtain a chromosome a 'shown in (3) of FIG.
[0017]
Similarly, chromosome c becomes “9450916734” by the exchange, and there are two 9 and 4 respectively, and there is no 8 and 2, so that 9 which originally existed in chromosome c has 8 Then, 4 is converted to 2 to obtain a chromosome c 'shown in (3) of FIG.
[0018]
The mutation operation unit 104 randomly selects a partial tour for the chromosome, and reverses the partial tour. For example, in the chromosome b shown in (3) in FIG. 15, the portion sandwiched between the triangular marks is reversed in order as shown in the explanation portion 112. In other words, the partial tour "3164" is reversed in the order of "4613" to obtain a chromosome b 'shown in (4) in FIG.
[0019]
The genetic operation of the fitness evaluation operation unit 101 to the mutation operation unit 104 is one generation, and the operations of the selection operation unit 102, the cross operation unit 103, and the mutation operation unit 104 are performed based on the parameters.
[0020]
After this one-generation operation, the procedure returns to the fitness evaluation operation unit 101 again, and the same operation is performed. This operation is repeated to perform the second-generation, third-generation,... Genetic operation.
As shown in FIG. 16, a standard arithmetic device using a bit string representation method suitable for a numerical function operation also includes an initial chromosome group generation unit 120, a fitness evaluation operation unit 121, a selection operation unit 122, and a cross operation unit 123. , A mutation operation unit 124 and the like.
[0021]
The necessary parameters are input to the initial chromosome population generation unit 120, and the parameters are selectively transmitted to the fitness evaluation operation unit 121, the selection operation unit 122, the crossover operation unit 123, and the mutation operation unit 124.
[0022]
The initial chromosome group generation unit 120 stochastically generates each chromosome a, b, c, d. For example, as shown in the group (1), chromosomes a = "1001010110", b = "1110101101", c = "0001100100", and d = "0110011000" are generated.
[0023]
The fitness evaluation calculation unit 121 calculates the fitness (a), (b), (c), and (d) of each chromosome a, b, c, and d. For example, in the case of a function operation, as shown in an explanation portion 130, a function F (x) is given in advance, and each of the chromosomes a, b, c, and d is used as a variable x, and a function F (x) corresponding thereto is used. Will be described as fitness values (a), (b), (c), and (d).
[0024]
When the 10-digit binary numbers of the chromosomes a, b, c, and d are represented by decimal numbers, they are "598", "941", "100", and "408", which are calculated by a given function F (x). Then, as shown in the explanation part 130 of FIG. 16, (a)>(b)>(c)> (d). Now, when finding a solution by the problem of finding the maximum value of the function F (x), the fitness of the chromosome a is the largest and the fitness of the chromosome d is the smallest.
[0025]
The selection operation unit 122 leaves the chromosomes in the population of the chromosomes a to d with high fitness, deletes those with low fitness, and if necessary deletes those with high fitness if necessary. Things. At this time, the remaining erasure can be selected by determining the probability by, for example, selecting a roulette wheel.
[0026]
In FIG. 16, (2) indicates that the fitness of the chromosomes a to d of the group (1) is calculated by the fitness evaluation calculation unit 121 as described above, It is shown that the chromosome d is deleted and the chromosome a is replicated as the chromosome d 'by determining that the chromosome d is the maximum and the fitness of the chromosome d is the minimum.
[0027]
The crossover unit 123 crosses two chromosomes to create two chromosomes of different types. For example, using the chromosomes a and c of the group {circle around (2)} in FIG. 16, crossover is performed between the second bit to the fifth bit from the head as shown in the explanation portion 131 as indicated by triangle marks. That is, chromosomes a 'and c' of the population (3) in FIG. 16 are obtained by exchanging these portions in chromosomes a and c.
[0028]
The mutation operation unit 124 randomly selects a bit for the chromosome, and inverts the bit of that part. For example, the second and fourth bits “1” and “0” of the chromosome b indicated by a triangle mark in (3) of FIG. b 'is obtained.
[0029]
The genetic operation of the fitness evaluation operation unit 121 to the mutation operation unit 124 is one generation, and the operations of the selection operation unit 122, the cross operation unit 123, and the mutation operation unit 124 are performed based on the parameters.
[0030]
After this one-generation operation, the procedure returns to the fitness evaluation operation unit 121 again, and the same operation is performed. This operation is repeated to perform the second-generation, third-generation genetic operation.
As described above, the standard GA for genetic manipulation using the symbol string representation method shown in FIGS. 15 and 16 generally includes an initial chromosome group generation stage, a fitness evaluation operation stage, a selection operation stage, and a crossover operation stage. , The operation can be divided into mutation operation stages. In each of these stages, the following operation is performed.
[0031]
In the initial chromosome population generation step, a symbol string representation (genotype) of the chromosome is stochastically generated as the initial chromosome population.
In the fitness evaluation calculation step, the symbol string expression is converted into the solution (phenotype) of the problem for each chromosome, and the fitness, that is, the superiority of the solution, is calculated for each phenotype.
[0032]
In the selection operation stage, the chromosome population is exchanged according to a certain rule such that more chromosomes with higher fitness are made more and chromosomes with lower fitness are less likely to survive.
[0033]
In the crossover operation step, in the chromosome group that has been replaced in the selection operation step, pairs are formed between chromosomes according to a certain probability, and a part of the symbol string expression is mutually converted. This operation aims at generating a further excellent chromosome by combining parts of the excellent chromosomes.
[0034]
In the mutation stage, there is a certain probability that a change is made to a part of the chromosome. This operation is intended to maintain the diversity of the chromosome population, which is inevitably reduced by the selection operation, and to generate different chromosomes that cannot be created by the crossover operation alone.
[0035]
Then, the generation from the fitness evaluation calculation stage to the mutation stage is regarded as one generation of genetic operation in GA, and this cycle (generation) is repeated until the chromosome population satisfies a certain censoring condition.
[0036]
By the way, GA has a feature that, as described above, it is possible to obtain an optimal solution or a suboptimal solution by a relatively simple mechanism of initial chromosome population generation, fitness evaluation, selection, crossover, and mutation. It is difficult to adopt a method appropriate for a given problem from among variations of a genetic operation and to adjust genetic operation parameters such as a crossover rate and a mutation rate.
[0037]
For example, as a method of evaluating the fitness, there is a method of simply setting the value of (a) to the fitness, a method of changing log (a) to the fitness, or a method of evaluating the other. As for the selection method, there is a method of simply leaving the first, second,... Rankings, and a method of roulette wheel selection.
[0038]
Regarding the crossover method, there are many choices such as selection of a chromosome to be crossed and selection of a crossing point. Mutation methods also have a number of choices, such as selection of a chromosome to be mutated, selection of a mutation position, and the like.
[0039]
Whatever method is used, it is sufficient to always go in the direction of optimization, but depending on the selection method, it may not go in the direction of a good solution. It is not known what method should be used to select parameters at each stage.
[0040]
Thus, in GA, for a given problem, an appropriate genetic manipulation method and a mathematical system for determining a set of operation parameters are still insufficient, and a huge number of At present it is necessary to rely on simulations.
[0041]
In order to evaluate the adopted genetic manipulation of GA using the simulation results, the following statistical analysis has been conventionally performed.
{Circle around (1)} The transition of the fitness in the chromosome population is plotted on a graph, and the rate of change of the average fitness (in the above example, the average value of (a) to (d)) is calculated, for example, to optimize the population. Look at the convergence to the solution or suboptimal solution. Thereby, the tendency of the change in the fitness is understood.
[0042]
{Circle around (2)} Before and after each genetic operation, the average and variance of the fitness in the population are calculated, and changes in the fitness due to each operation are observed. This shows how much the fitness in the population has changed before and after selection. In addition, if the fitness is calculated after a genetic operation, for example, crossover, the effect of the calculation is known. As described above, if the fitness is calculated for each step, a change in the fitness can be found.
[0043]
{Circle around (3)} In the function optimization problem and the like, look at the distribution of phenotypes of chromosomes in the solution space.
(4) See the distribution of alleles at each locus, that is, the distribution of the types of symbols that can be placed on one point of the symbol string of this locus, with each chromosome having 10 loci in FIGS. .
[0044]
As described above, the operating principle of GA is to convert some points in the search space into a symbolic expression form called chromosome, and gradually perform symbolic operations on these chromosome populations as genetic operations, Is to find the optimal solution.
[0045]
GA has a theoretical base that is undeveloped due to its characteristics such as handling populations, genetic operations are discrete / non-linear, and the number of parameters to be tuned is large. Therefore, in order to make the search by GA efficient, it is necessary to rely on empirical rules based on a large amount of simulation, trial and error, and the like.
[0046]
Therefore, the present inventors have analyzed and visualized the transition of the distribution of GA chromosome populations graphically and analytically, so that analysis and visualization display that makes it easy to grasp and understand the behavior and generation change of chromosome populations The devices were invented as Japanese Patent Application No. 7-238389, "Analysis and Display Processing Apparatus for Genetic Algorithm", and Japanese Patent Application No. 8-123161, and filed a patent application.
[0047]
[Problems to be solved by the invention]
In these inventions, GA execution results are graphically displayed by various analysis methods and display methods. However, these diagrams only show the chromosome population of GA from a certain point of view, and it is difficult to correctly understand all the behavior of the chromosome population, and it is difficult to judge comprehensively from various analysis results. Is required.
[0048]
In order to understand from a multifaceted viewpoint, it is necessary to determine the location of the same chromosome on each of the different views, or what kind of population distribution in the same generation, etc. It is important to know the correspondence between the two.
[0049]
Accordingly, an object of the present invention is to visually indicate the correspondence between the same chromosome, the same generation, and the same execution event in a plurality of types of display diagrams on the display diagram, thereby enabling the behavior of the chromosome population and generation change. An object of the present invention is to provide a genetic algorithm analysis display processing device that facilitates understanding.
[0050]
[Means for Solving the Problems]
FIG. 1 shows a principle diagram of the present invention for achieving the above object. In FIG. 1, 1 is an image display device, 1-0 is a display screen, 2 is a position designation device, 3 is an auxiliary storage device, 4 is a character input device, 5 is a control unit, 6-1 is a first display unit, 6 -2 is a second display unit, and 6-n is an n-th display unit. The present invention is configured as follows.
[0051]
(1) The distribution of chromosomes 1, 2,..., 10 for a specific generation of the chromosome population of the genetic algorithm is displayed on the display screen 1-0 of the image display device 1, for example, in a cluster dendrogram A and a variable domain. As shown in the distribution diagram B, a plurality of types of display diagrams are output, and when a specific portion of one of the display diagrams, such as chromosome 1, is designated, a corresponding portion of another diagram corresponding to the specific portion, such as the other, is displayed. Is distinguished by, for example, color or blinking.
[0052]
(2) Instead of a plurality of types of display diagrams for the specific generation, a generation change diagram of a fitness distribution of a chromosome population from a certain generation to a certain generation and a generation change diagram of another distribution of a chromosome variable 1 are displayed. In addition, a feature of the chromosome designated on one display diagram is distinguishably displayed on the other display diagram.
[0053]
(3) One of a display diagram for a specific generation such as a cluster dendrogram of a chromosome population and a display diagram for a generation range such as a generation change diagram of a chromosome population fitness distribution. The chromosome designated by the above is distinguished and displayed on the other display diagram.
[0054]
(4) Image display of the amount of analysis in each generation of the chromosome group of the genetic algorithm, such as the generation change diagram of the maximum fitness and the generation change locus diagram of the fitness maximum value and the fitness deviation, for a range from a certain generation to a certain generation By designating a certain generation on one display diagram by the position designation device 2 among a plurality of types of display diagrams illustrated on the display screen 1-0 of the device 1, other image display devices on the image display device 1 are displayed. It is characterized in that corresponding generations on the display diagram are distinguishably displayed.
[0055]
(5) On the display screen , A chromosome for a specific generation or range of generations Display a generation change diagram of a certain distribution of chromosome populations of a certain range of generations. , One When the generation information display portion of the figure is designated by the position specifying means, the corresponding generation information display portion on another display diagram is displayed in a distinctive manner. It is characterized by the following.
[0056]
(6) On the display screen, a plurality of For multiple execution cases of the genetic algorithm executed under different conditions, is there Graphical analysis volume A distribution diagram of a certain analysis amount in a different execution case and a distribution diagram of another analysis amount in a different execution case illustrating another type of analysis amount of the same execution case are displayed. Of certain conditions Execution example When the display part of is designated by the position specifying means, another distribution Corresponding execution example on the diagram The display part of It is distinguished and displayed.
[0057]
(7) A plurality of types of display examples of execution examples for a generation range are displayed, such as a generation change diagram of the best chromosome and a generation diagram of fitness deviation, and the execution example designated on one display diagram is used for another. Are distinguished on the display diagram of FIG.
[0058]
(8) One of a display diagram between a display diagram for a specific generation such as a distribution diagram of variable domain values in different execution cases and a display diagram for a generation range such as a diagram showing a change in generation of the best chromosome. Is characterized in that the execution examples specified by the above are distinguishably displayed on another display diagram.
[0059]
(9) A plurality of types of display diagrams of execution examples for a certain generation range are displayed, such as a generation change diagram of the best chromosome and a generation change diagram of the fitness deviation, and according to the generation indicated in one of the display diagrams. In this case, all execution cases belonging to the corresponding generation on the other display diagram are distinguishably displayed.
[0060]
According to the present invention, the following functions and effects are obtained corresponding to the above (1) to (9).
(1) The positional relationship of the same chromosome on a plurality of types of display diagrams can be easily recognized, thereby helping to understand the characteristics of the chromosome population distribution.
[0061]
(2) The positional relationship of the same chromosome on a plurality of types of display diagrams can be easily recognized, thereby helping to understand the characteristics of the chromosome population distribution.
(3) The positional relationship of the same chromosome on a plurality of types of display diagrams can be easily recognized, thereby helping to understand the characteristics of the chromosome population distribution.
[0062]
(4) The positional relationship between the amounts of analysis of the same generation on a plurality of types of display diagrams can be easily recognized, thereby helping to understand the characteristics of generational changes in the amount of analysis of a chromosome population.
(5) The positional relationship between the amounts of analysis of the same generation on a plurality of types of display diagrams can be easily recognized, thereby helping to understand the characteristics of generational changes in the amount of analysis of a chromosome population.
[0063]
(6) It is possible to easily recognize the positional relationship between the amounts of analysis of the same execution case on a plurality of types of display diagrams, thereby helping to understand the characteristics of the difference between the execution cases under different conditions.
[0064]
(7) It is possible to easily recognize the positional relationship between the amounts of analysis of the same execution case on a plurality of types of display diagrams, thereby helping to understand the characteristics of the difference in the generation change of the execution case under different conditions. .
[0065]
(8) The positional relationship between the amounts of analysis of the same execution case on a plurality of types of display diagrams can be easily recognized, thereby helping to understand differences between execution cases under different conditions.
(9) Differences in the amount of analysis of execution cases in the same generation on a plurality of types of display diagrams can be easily recognized, thereby helping to understand differences in execution cases under different conditions.
[0066]
BEST MODE FOR CARRYING OUT THE INVENTION
An embodiment of the present invention will be described with reference to FIGS. FIG. 2 is a diagram of an embodiment of the present invention, FIG. 3 is a cluster tree diagram of a chromosome population, FIG. 4 is a variable domain distribution map of the chromosome population, FIG. 5 is a generation change diagram of a chromosome population fitness distribution, FIG. Is a generation change diagram of the distribution of the value of the variable 1, FIG. 7 is a generation change diagram of the maximum fitness, FIG. 8 is a generation change locus diagram of the fitness maximum value and the fitness deviation, and FIG. 9 is fitness in different execution examples. 10 is a distribution diagram of variable domain values in different execution cases, FIG. 11 is a generation change diagram of the best chromosome, and FIG. 12 is a generation change diagram of the fitness deviation.
[0067]
In FIG. 2, the same symbols as those in the other figures denote the same parts, 1 is an image display device, 1-0 is a display screen of the image display device 1, 2 is a position designation device, 3 is an auxiliary storage device, 4 is a character input device, Reference numeral 5 denotes a control unit, 6-1 denotes a first display unit, 6-2 denotes a second display unit, and 6-n denotes an n-th display unit. 7-1, 7-2, 7-n are a first display unit, a second display unit, a display diagram indicating unit transmitting means of the n-th display unit, and 8-1, 8-2, 8-n are a first display unit. , 9-2, 9-n are the first display unit, the second display unit, the display diagram creation unit of the n-th display unit, 10-1, 10-2, and 10-n are first display units, second display units, and display diagram indicator detecting means of the nth display unit, and 11-1, 11-2, and 11-n are first display units. , A second display unit, a display diagram indicating means of an n-th display unit, 12-1, 12-2, and 12-n are display diagram control means of a first display unit, a second display unit, and an n-th display unit; 13-1, 13-2, and 13-n are storage units of the first display unit, the second display unit, and the n-th display unit.
[0068]
The image display device 1 displays, on its display screen 1-0, a cluster tree diagram of the chromosome population, a variable domain distribution map of the chromosome population, a generation change diagram of the chromosome population fitness distribution, and the like, for example, as shown in FIGS. The figure as shown is displayed. The display unit that displays the display screen 1-0 may be constituted by, for example, a projector in addition to a display such as a CRT or a liquid crystal.
[0069]
The position designation device 2 designates the position of the display screen 1-0, and is constituted by, for example, a mouse.
The auxiliary storage device 3 stores chromosome groups for each generation created and evaluated by the method shown in FIGS. 15 and 16 and various figures such as the cluster tree diagrams shown in FIGS. It is stored. Note that the means for creating these various figures is not a characteristic part of the present invention, and will not be described for simplicity.
[0070]
The character input device 4 is for inputting designation of a generation, a command, and the like, and is constituted by, for example, a keyboard.
The control unit 5 performs a genetic algorithm operation process as shown in FIGS. 15 and 16 by a genetic algorithm operation device (not shown), and stores necessary data among the obtained data in the auxiliary storage device 3, Based on these data, various figures as shown in FIGS. 3 to 12 are created based on drawing means not shown, and these are stored in the auxiliary storage device. The position information specified by the position specifying device 2 is Control such as transmission to the display unit of the figure is performed.
[0071]
The first display unit 6-1 to the n-th display unit 6-n create a screen to be displayed on the display screen 1-0 based on the result of the genetic algorithm operation processing. For example, the first display unit 6-n 1 creates a cluster tree diagram as shown in FIG. 3, and the second display unit 6-2 creates a variable domain distribution map of the chromosome group as shown in FIG.
[0072]
The first display unit 6-1 includes a display diagram indicating object transmitting unit 7-1, a display diagram indicating unit receiving unit 8-1, a display diagram creating unit 9-1, a display diagram indicating unit detecting unit 10-1, and a display diagram indicating unit. It comprises an object illustration unit 11-1, a display diagram control unit 12-1, a storage unit 13-1, and the like.
[0073]
Similarly, the second display unit 6-2 includes a display diagram indicating object transmitting unit 7-2, a display diagram indicating unit receiving unit 8-2, a display diagram creating unit 9-2, a display diagram indicating unit detecting unit 10-2, and a display diagram. An n-th display unit 6-n includes a display diagram indicating unit transmitting unit 7-n, a display diagram indicating unit receiving unit 8, and an indicator indicating unit 11-2, a display diagram controlling unit 12-2, and a storage unit 13-2. -N, a display diagram creation unit 9-n, a display diagram indicator detection unit 10-n, a display diagram indicator illustration unit 11-n, a display diagram control unit 12-n, and a storage unit 13-n.
[0074]
The display diagram indicating object transmitting means 7-1, when the own screen (display area of the cluster tree diagram in the case of the first display unit 6-1) on the display screen 1-0 is indicated by the position specifying device 2, This instruction is transmitted to the other second display units 6-2 to n-th display units 6-n via the control unit 5.
[0075]
When the position designated by the position designation device 2 is another screen, the display diagram designation receiving means 8-1 transmits the information indicating the designated one transmitted from the other display unit via the control unit 5. And receive it.
[0076]
The display diagram creating means 9-1 analyzes the result of the genetic algorithm operation processing and creates diagram information based on the display diagram as shown in FIG.
The display diagram indicator detecting means 10-1 receives the position information from the position specifying device 2 via the control unit 5, and determines the specific information of the position specified portion, for example, which chromosome, generation, execution example, and the like. It is to ask whether it was instructed. For this reason, the data of the display diagram is stored in the storage means 13-1, and what is instructed is determined by referring to the data.
[0077]
The display diagram indicating object indicating means 11-1 performs, on the self-display diagram, the specified object transmitted from another display unit, for example, based on a chromosome, a generation, an execution example, or the like. A display having a visual effect such as a color, a pattern, a blink, and an arrow graphic, that is, a distinctive display is performed.
[0078]
The display diagram control unit 12-1 includes a display diagram instruction transmitting unit 7-1, a display diagram instruction receiving unit 8-1, a display diagram creating unit 9-1, a display diagram instruction detecting unit 10-1, and a display diagram instruction. It selectively controls the object illustration means 11-1 and transmits and receives data from the control unit 5, the image display device 1, the position designation device 2, and the like.
[0079]
Display diagram indication transmitting means 7-2, display diagram indication receiving means 8-2, display diagram creating means 9-2, display diagram indication detecting means 10-2, display diagram indication of the second display unit 6-2. The display unit 11-2, the display control unit 12-2, the storage unit 13-2, and the display instruction transmitting unit 7-n, the display instruction receiving unit 8-n, and the display diagram of the n-th display unit 6-n. The creation unit 9-n, the display diagram indicator detection unit 10-n, the display diagram indicator illustration unit 11-n, the display diagram control unit 12-n, and the storage unit 13-n also have the same names as those of the first display unit 6-1. The same operation as is performed.
[0080]
Prior to describing the operation of the present invention, an example of an optimization problem and a chromosome population used in the description of the present invention will be described.
In the problem of maximizing the integer function of two integer variables, the domain of variable 1 and variable 2 are both {0, 1, 2,... 255}, and the range of the function is {0, 1, 2,.・ It shall be 1023 °.
[0081]
The chromosome represents the value of each variable as a binary value (up to 255, 8 bits in length), and has a total length of 16 bits. Table 1 shows an example of a chromosome population consisting of 10 members. The function value is used as the fitness of the chromosome.
[0082]
[Table 1]

[0083]
Table 1 is briefly described. The chromosome No. 1 is composed of “75” for variable 1 and “204” for variable 2, indicating that the fitness is “993”. The binary numbers “75” and “204” are “01001011” and “11001100” displayed as chromosomes.
[0084]
A. Explanation of display diagram
(1) Cluster dendrogram of chromosome population
3 to 12 which are display diagrams used in the description of the present invention will be described.
[0085]
FIG. 3 is a cluster dendrogram of the chromosome population shown in Table 1. Generally, between chromosomes of a genetic algorithm, an amount called similarity representing “similar chromosomes” and “similar chromosomes” can be defined. Further, a group of chromosomes having a certain similarity value or less is called a cluster. One chromosome can also be regarded as a special cluster. In general, similar similarities can be defined between clusters.
[0086]
As shown in FIG. 3, the cluster dendrogram has a similarity between chromosomes on the horizontal axis and chromosomes on the vertical axis. Multiple clusters or chromosome cl ₁ , Cl ₂ , Cl ₃ .. Are D, the horizontal line from the cluster on the vertical axis is connected by a vertical line when the similarity value on the horizontal axis is D. FIG. 3 shows an example in which the similarity between chromosomes 6 and 7 is 4, and the similarity between chromosomes 8 and 10 is 5. The cluster dendrogram is effective for knowing the degree of variation in a chromosome population. The method of creating a cluster dendrogram of a chromosome population has been described in the above-mentioned Japanese Patent Application No. 238389 by the present inventors. Further, in the present invention, it is only necessary that the chromosome or the cluster or generation of the chromosome on the display diagram can be designated by a position designation device such as a mouse.
[0087]
(2) Chromosome population variable domain distribution map
As shown in FIG. 4, chromosomes 1, 2,... Are represented by, for example, two variables. For example, as shown in Table 1, when chromosomes 1, 2, 3,... Are represented by 16 bits, variable 1 is indicated by a decimal number indicated by the first 8 bits, and variable 2 is indicated by a decimal number indicated by the subsequent 8 bits. A variable domain distribution map as shown in FIG. 4 is created by combining the variables 1 and 2. Generally, a chromosome is represented by n variables, two appropriate variables are selected from the n variables, and a two-dimensional domain spanned by the two variables is defined as the X and Y axes corresponding to the variable values represented by each chromosome. The position is indicated by a mark figure.
[0088]
For example, when the chromosome population of Table 1 is displayed as shown in FIG. 4, this suggests the presence of a combination of variables showing some peak near the region surrounding chromosomes 2, 3, 7, and 9, , 9 near the region surrounding the chromosomes 5 and 6, and the presence of a combination of variables showing the same peak near the regions surrounding chromosomes 5 and 6.
[0089]
(3) Generation change diagram of chromosome population fitness distribution
As shown in FIG. 5, the horizontal axis represents generations, and the vertical axis represents fitness, and all chromosomes in the population at each generation are represented by dots, and each point represents one chromosome.
[0090]
FIG. 5 shows how the fitness of chromosomes changes every time 10 chromosomes in 0 generation go through 1 generation, 2 generations... The fitness of the chromosome changes depending on the calculation result of the genetic algorithm, and may be equal. In one generation, there is an overlap in a part of the fitness of the chromosome, and in the 14th generation and thereafter, the fitness completely overlaps. As shown in the above-mentioned Japanese Patent Application No. 7-238389, chromosomes may be connected by lines between adjacent generations, but the presence or absence of the connection is not important to the present invention, and is omitted in FIG.
[0091]
(4) Generation change diagram of distribution of variable 1 value
As shown in FIG. 6, the horizontal axis represents the generation, and the vertical axis represents the value of variable 1 represented by the chromosome (for example, the first 8-bit decimal number of the chromosome in Table 1). Is indicated by dots. As a result, the generation change of the distribution of the value of the variable 1 within the group as well as the generation change, that is, the convergence state can be indicated.
[0092]
(5) Generation change diagram of maximum fitness
As shown in FIG. 7, the horizontal axis shows generations and the vertical axis shows fitness of chromosomes, and the maximum value of the fitness of chromosomes in the population at each generation is indicated by dots. For example, as shown in FIG. 5, when the fitness of the chromosome in the population at each generation is indicated, the maximum value of the fitness at each generation is displayed in FIG. As a result, it is possible to show how the optimal solution is being found through generations and how the generation changes.
[0093]
(6) Generation change locus diagram of fitness maximum value and fitness deviation
As shown in FIG. 8, the horizontal axis indicates the maximum value of the fitness within the chromosome population, and the vertical axis indicates the fitness standard deviation value. . As a result, it is possible to show the state of the generation change in the form of a locus in the distribution state of the fitness level after generations.
[0094]
(7) Distribution chart of fitness values in different execution cases
As shown in FIG. 9, the horizontal axis represents the maximum fitness value within the group, the vertical axis represents the fitness standard deviation value, and the fitness at a specific generation is shown for execution cases executed under several different conditions. The maximum value and the deviation value are indicated by dots. FIG. 9 shows six execution examples in which the initial values of the pseudo random numbers are different. As a result, it is possible to show the difference in the result due to the difference in the condition.
[0095]
(8) Variable domain value distribution diagram in different execution examples
As shown in FIG. 10, with the two-dimensional domain spanned by a certain two variables as the X and Y axes, the best chromosome in the population, that is, the fitness in the population, for a specific generation, for an execution example under different conditions This is a point display of the variable value of the chromosome with the largest value. Thereby, for example, a difference in the result due to a difference in the initial value of the pseudo random number can be indicated.
[0096]
(9) Generation change diagram of the best chromosome
As shown in FIG. 11, the horizontal axis shows generations and the vertical axis shows fitness values of chromosomes. For some execution examples of genetic algorithms, the best chromosome in each generation, that is, the chromosome having the largest fitness in the population, is plotted. Displayed and connected. Each point represents one chromosome. The maximum fitness values of the execution cases 1 and 2 overlap in the eleventh generation, and the maximum fitness values of the execution cases 1 and 3 overlap in the 14th generation. This makes it possible to know the difference between execution cases calculated under different conditions.
[0097]
(10) Generation change diagram of fitness deviation
As shown in FIG. 12, the horizontal axis represents the generation and the vertical axis represents the standard deviation of the fitness values of the entire chromosome population, and for some execution examples of genetic algorithms, the standard deviation of the fitness value of the entire population at each generation. Are displayed as dots and connected. In the 16th generation, the standard deviation values of execution examples 1 and 2 overlap. As a result, it is possible to show the state of the generation change in the degree of dispersion of the fitness value, that is, the variation of the group, after passing through the generations.
[0098]
B, description of operation of the present invention
(1) Description of Operation of the Present Invention (Part 1)
In FIG. 2, for example, an operation result by the genetic algorithm described with reference to FIG. 15 is stored in the auxiliary storage device 3. When the display command is input from the character input device 4, the first display unit 6-1 displays the display diagram creating means 9-1 on the basis of the calculation result stored in the auxiliary storage device 3 as shown in FIG. A diagram is created and sent to the control unit 5 via the display diagram control means 12-1.
[0099]
The control unit 5 transmits the cluster tree diagram to the image display device 1 which displays the cluster tree diagram at a predetermined position on the display screen 1-0. The image display device 1 reduces the size of the display screen 1-0 to a size predetermined for the first display unit 6-1 and adds the coordinates and the figure number of the display screen 1-0 to the area. , The figure number, for example, G1 is added and displayed. The cluster tree diagram created by the display diagram creating unit 9-1 is separately stored in the storage unit 13-1.
[0100]
In the second display unit 6-2, the display diagram creating means 9-2 creates a variable domain distribution map of the chromosome group as shown in FIG. Then, the image display device 1 displays a figure number, for example, G2 in a predetermined area of the display screen 1-0. The variable domain distribution map of the chromosome group is separately stored in the storage means 13-2.
[0101]
In this way, each display unit operates, and each figure is displayed on the display screen 1-0.
Next, when the operator decides to observe the relationship between the cluster tree state of the chromosome group and the variable domain distribution state of the chromosome group, the display screen selection signals SG1 and SG2 are input from the character input device 4. Thereby, the control unit 5 transmits this to the display diagram control unit 12-1 of the first display unit 6-1 and the display diagram control unit 12-2 of the second display unit 6-2 to set the selected display operation state. If there is no input of a display screen selection signal from the character input device 4, all display diagrams are to be selected.
[0102]
Then, when the operator designates the mark figure of the chromosome 10 (see FIG. 3) of the cluster tree diagram of the figure number G1 on the display screen 1-0 by the position designation device 2 such as a mouse, the position is detected by the image display device 1. Is transmitted to the control unit 5. The control unit 5 converts the position into the original coordinate position signal of the cluster tree diagram shown in FIG. 3, that is, the original coordinate position signal transmitted from the display diagram control unit 12-1 of the first display unit 6-1. This original coordinate position signal is transmitted to the display diagram control means 12-1 of the first display unit 6-1. Note that the conversion to the original coordinate position signal can be performed not on the control unit 5 but on the display diagram side.
[0103]
The display diagram control means 12-1 of the first display unit 6-1 accesses the storage means 13-1 based on the original coordinate position signal and recognizes that this is the chromosome 10. Then, it instructs the display diagram indicating object transmitting means 7-1 to notify the other display units that the chromosome 10 is to be distinguished and displayed. Accordingly, the display diagram indicating object transmitting unit 7-1 transmits a transmission signal instructing the second display unit 6-2 to the n-th display unit 6-n, which are other display units, to display the chromosome 10 in a distinguished manner. It is transmitted via the control unit 5. Note that the recognition can be performed by the display diagram indicating object transmitting means 7-1.
[0104]
At this time, since only the second display unit 6-2 is in the operating state, this transmission signal is received by the display diagram indication receiving unit 8-2 via the display diagram control unit 12-2, and the received signal is received. Is decoded, and the display diagram control means 12-2 instructs the display diagram indicating object indicating means 11-2 to distinguish and display the chromosome 10.
[0105]
As a result, the display diagram indicating unit 11-2 accesses the storage unit 13-2 to read out the position of the chromosome 10 in the variable domain distribution diagram of the chromosome group shown in FIG. This is transmitted to the control unit 5 via the display diagram control means 12-2. In addition, the display diagram indicating means 11-2 can form the entire display diagram which is distinguished and displayed in cooperation with the display diagram creating means 9-2, and can display it on the image display device via the control unit 5. .
[0106]
Thereby, the control unit 5 calculates the position of the chromosome 10 in the figure number G2 on the display screen 1-0, and transmits the calculated position to be distinguished and displayed to the image display device 1, and The mark figure of the chromosome 10 at the number G2 is displayed in a distinguishing manner, for example, blinking, being displayed in a different color, being patterned, or being accompanied by an arrow figure.
[0107]
In this way, the operator can observe the display state of the chromosome on the diagram number G2 corresponding to the chromosome of the diagram number G1, and can observe the correspondence between the cluster tree state and the variable domain distribution state.
[0108]
Similarly, when the horizontal line indicating the cluster consisting of chromosomes 1, 8, and 10 on the cluster dendrogram is designated by the position designation device 2, the mark figure of chromosome numbers 1, 8, and 10 on the variable definition distribution diagram of the chromosome population is obtained. Are all displayed separately.
[0109]
Conversely, if the mark figure of chromosome 4, for example, is designated by the position designation device 2 on the variable domain distribution diagram of the chromosome group on the figure G2, the chromosome 4 on the cluster dendrogram on the figure G1 is similarly designated. Are markedly displayed.
[0110]
In addition, by designating, for example, mark shapes of chromosome numbers 6 and 7 on the variable domain distribution map of the chromosome group on the diagram number G2 with a diagram such as a circle, the chromosomes on the cluster dendrogram on the diagram number G1 are displayed. The mark figures of Nos. 6 and 7 are distinguished and displayed.
[0111]
Generally, in the cluster tree diagram or the variable domain distribution map of the chromosome group, each chromosome is not always numbered. 3 and 4, the numbers are described for convenience of explanation.
[0112]
(2) Description of Operation of the Present Invention (Part 2)
This is a case where a generation change diagram of the chromosome population fitness distribution (FIG. 5) and a generation change diagram of the distribution of the value of the variable 1 of the chromosome (FIG. 6) are used from the display diagram displayed on the display screen 1-0.
[0113]
For example, when a certain chromosome on the generation change diagram of the eighth generation fitness distribution, for example, chromosome 5 (the number is omitted in FIG. 5) is designated by the position specifying device 2, the third display unit (not shown) is provided in the same manner as described above. A transmission signal for instructing to distinguish and display the eighth generation chromosome 5 is output from the display diagram indicating object transmitting means.
[0114]
This is received by the display indicator indicating means of the fourth display unit (not shown) for creating the display diagram of FIG. 6, and the generation change diagram of the distribution of the value of the variable 1 of the chromosome on the display screen 1-0 is shown in FIG. The value of the variable 1 of the chromosome 5 of the generation is displayed in a distinguishing manner, for example, blinking.
[0115]
Conversely, when the chromosome 5 of the eighth generation in the generation change diagram of the distribution of the value of the variable 1 on the display screen 1-0 is designated by the position specifying device, the eighth generation chromosome on the generation change diagram of the chromosome group fitness distribution is displayed. The mark figure corresponding to the chromosome 5 is displayed.
[0116]
In this way, the correspondence between the distribution of the value of the variable 1 on the domain and the distribution of the fitness on the range can be known.
(3) Description of Operation of the Present Invention (Part 3)
In this case, a cluster tree diagram of a chromosome group (FIG. 3) and a generation change diagram of a chromosome group fitness distribution (FIG. 5) are used from the display diagram displayed on the display screen 1-0. For example, when the chromosome 5 on the cluster dendrogram of the chromosome population of the eighth generation is designated by the position specifying device 2, the display diagram indicator transmitting means 7-1 of the first display unit 6-1 transmits the chromosome 5 in the same manner as described above. A transmission signal instructing that the chromosomes 5 of the eight generations be displayed in a distinguished manner is output.
[0117]
This is received by the display indicator indicating means of the third display unit (not shown) for creating the display diagram of FIG. 5, and the chromosome of the eighth generation of the generation change diagram of the chromosome population fitness distribution on the display screen 1-0 is displayed. The mark figure corresponding to No. 5 is distinguished and displayed.
[0118]
When a horizontal line indicating a cluster composed of chromosomes 6 and 7 on the cluster dendrogram of the chromosome group is designated by the position specifying device 2, all the mark figures corresponding to chromosomes 6 and 7 in the generation change diagram of the chromosome group fitness distribution are displayed. It is distinguished and displayed.
[0119]
Conversely, when the mark design of the sixth generation chromosome 4 in the generation change diagram of the chromosome population fitness distribution on the display screen 1-0 is designated by the position designation device 2, for example, the cluster tree diagram of the sixth generation chromosome population The mark figure of the upper chromosome 4 is distinguished and displayed. At this time, when the cluster tree diagram of the chromosome group displayed on the display screen 1-0 is of another generation, for example, the eighth generation, the following control is performed.
[0120]
That is, when the display diagram indicator receiving means 8-1 of the first display unit 6-1 receives a signal instructing the chromosome 4 of the sixth generation to be distinguished and displayed, the signal is decoded. At this time, the display diagram control means 12-1 recognizes that the one being displayed on the display screen 1-0 is the eighth generation and is not the sixth generation, so the control unit 5 is provided with an auxiliary storage device. The third to sixth generation operation results are read and transmitted.
[0121]
Then, the display diagram creating means 9-1 creates a cluster tree diagram of the chromosome population of the sixth generation, sends it to the image display device 1 via the control unit 5, and then displays the cluster tree diagram on the display screen 1-0. A cluster dendrogram of the chromosome population of six generations is displayed. At this time, this is stored in the storage means 13-1.
[0122]
Subsequently, the display diagram creating means 9-1 indicates the position of the corresponding chromosome 4 in the sixth generation cluster dendrogram to the display diagram control means 12-1, so that the display diagram indicating object indicating means 11-1 performs this processing. Is instructed to the control unit 5 to distinguish and display. The control unit 5 calculates the position of the chromosome 4 and instructs the image display device 1 to distinguish and display the position. In this way, even if the one being displayed and another generation are designated, the designated one can be distinguished and displayed.
[0123]
In this way, it is possible to know the correspondence between the cluster tree diagram and the generation change diagram of the chromosome population fitness distribution.
(4) Description of Operation of the Present Invention (Part 4)
In this case, a generation change diagram of the maximum fitness value (FIG. 7) and a generation change locus diagram of the fitness maximum value and the fitness deviation (FIG. 8) are used from the display diagram displayed on the display screen 1-0. For example, when the position designating device 2 designates the third generation mark graphic in the generation change diagram of the maximum fitness, the third generation is distinguished from the display diagram indicator transmitting means of the fifth display unit (not shown) in the same manner as described above. A transmission signal for instructing display is output, and this is received by the display diagram indication receiving means of the sixth display unit (not shown) for creating the display diagram of FIG. 8, and the fitness on the display screen 1-0 is maximized. The mark figure corresponding to the third generation of the generation change locus diagram of the value and the fitness deviation is distinguished and displayed.
[0124]
Conversely, when the sixth designating mark graphic on the generation change locus diagram of the maximum fitness value and the fitness deviation is designated by the position specifying device 2, the sixth generation mark graphic of the maximum fitness generation change diagram is distinguished and displayed. Is done.
[0125]
Thus, the generation change state of the maximum fitness and the generation change trajectory state of the fitness can be observed.
(5) Description of operation of the present invention (part 5)
In this case, a generation change diagram of the chromosome population fitness distribution (FIG. 5) and a generation change locus diagram of the fitness maximum value and the fitness deviation (FIG. 8) are used from the display diagram displayed on the display screen 1-0. For example, when the seventh generation is designated by the position specification device 2 on the coordinate scale of the generation change diagram of the chromosome population fitness distribution, the display designation indicator transmission means of the third display unit (not shown) transmits the seventh generation in the same manner as described above. A transmission signal for instructing the seven generations to be displayed is output. The transmission signal is received by the display instruction receiving means of the sixth display unit (not shown) for creating the display of FIG. 8, and is displayed on the display screen 1-0. The mark figure corresponding to the seventh generation of the generation change locus diagram of the fitness maximum value and the fitness deviation of the fitness difference is displayed.
[0126]
Conversely, when the fourth generation mark graphic on the generation change locus diagram of the fitness maximum value and the fitness deviation is designated by the position specifying device 2, a display diagram designation on the sixth display unit (not shown) is made in the same manner as described above. A transmission signal for instructing the fourth generation to be distinguished and displayed is output from the object transmitting means, and the transmission signal is received by the display figure indicating object receiving means of the third display unit (not shown) for creating the display diagram of FIG. The fourth generation is discriminated and transmitted to the display figure / pointer indicating means for distinguishing and displaying the fourth generation.
[0127]
Accordingly, the display diagram indication receiving means accesses the storage device, reads out the positions of all the chromosomes of the fourth generation, and transmits to the control unit 5 via the display control means that the chromosomes are distinguishably displayed.
[0128]
Based on this, the control unit 5 calculates the positions of all chromosomes of the fourth generation as positions on the display screen 1-0, and transmits these to the image display device 1. In this way, the mark figures of all the chromosomes belonging to the fourth generation in the generation change diagram of the chromosome group fitness distribution are distinguished and displayed.
[0129]
In this manner, it is possible to observe the generation change of the chromosome group fitness distribution and the generation change trajectory state of the fitness maximum value and fitness deviation.
(6) Description of Operation of the Present Invention (Part 6)
In this case, a distribution diagram of fitness values in different execution cases (FIG. 9) and a variable domain value distribution diagram (FIG. 10) in different execution cases are used from the display diagram displayed on the display screen 1-0. For example, when the mark figure of the execution example 2 is designated by the position specifying device 2 on the distribution diagram of the fitness values in the different execution cases of the specific generation, the display diagram designation of the seventh display unit (not shown) is performed in the same manner as described above. A transmission signal for instructing the execution example 2 to be distinguished and displayed is output from the object transmission means.
[0130]
This is received by the display instruction indicator receiving means of the eighth display unit (not shown) for creating the display diagram of FIG. 10, and the variable domain values in the same specific generation and different execution examples on the display screen 1-0 are displayed. The mark figure corresponding to the execution example 2 of the distribution diagram is distinguished and displayed.
[0131]
Conversely, when the mark figure of execution case 3 on the variable domain value distribution diagram (FIG. 10) in different execution cases of a specific generation is designated by the position specifying device 2, the fitness value of the same specific generation in different execution cases The mark figure corresponding to the execution example 3 on the distribution diagram (FIG. 9) is distinguished and displayed.
[0132]
As a result, it is possible to observe the distribution of fitness values in different execution cases and the distribution of variable domain values in different execution cases.
(7) Description of operation of the present invention (part 7)
This is a case where the generation change diagram of the best chromosome (FIG. 11) and the generation change diagram of the fitness deviation (FIG. 12) are used from the display diagram displayed on the display screen 1-0. For example, when a white circle portion of a proviso or a character portion of "execution example 1" displayed in a rectangular dotted frame on the generation change diagram of the best chromosome is designated by the position designation device 2, the ninth display unit ( A display signal (not shown) outputs a transmission signal for instructing the execution example 1 to be displayed in a distinctive manner.
[0133]
This is received by the display diagram indicator receiving means of the tenth display unit (not shown) for creating the display diagram of FIG. 12, and corresponds to the execution example 1 of the generation change diagram of the fitness deviation on the display screen 1-0. The mark figure and its connection line are displayed separately.
[0134]
Conversely, when the position specification device 2 designates the black circle portion of the proviso portion or the character portion of "executing example 2" displayed within the rectangular dotted frame on the generation change diagram of the fitness deviation (FIG. 12), the maximum fitness The mark figure corresponding to the execution example 2 on the generation change diagram (FIG. 11) and its connection line are displayed in a distinguished manner.
[0135]
It is to be noted that the instruction by the position specifying device 2 may indicate a connection line even if it is not in the above proviso part.
Thereby, it is possible to observe the maximum fitness generation change state and the fitness change generation change state in the same execution case.
[0136]
(8) Description of Operation of the Present Invention (Part 8)
This is a case where a variable domain distribution map (FIG. 10) and a generation change diagram of the best chromosome (FIG. 11) in different execution cases are used from the display diagram displayed on the display screen 1-0. For example, when the mark figure of the execution example 1 is designated by the position specifying device 2 on the variable domain value distribution diagram in the different execution cases of the fourth generation, the display diagram of the eighth display unit (not shown) is made in the same manner as described above. A transmission signal for instructing the fourth generation execution case 1 to be displayed in a distinctive manner is output from the pointer transmitting means.
[0137]
This is received by the display diagram indicator receiving means of the ninth display unit (not shown) for creating the display diagram of FIG. 11, and is displayed in the fourth generation of the execution example 1 of the best chromosome generation change diagram on the display screen 1-0. Corresponding mark figures are distinguished and displayed.
[0138]
Conversely, if the mark figure of the fifth generation execution case 3 on the best chromosome generation change diagram (FIG. 11) is designated by the position designation device 2, the variable domain value distribution diagram in the fifth execution different execution case The mark figure corresponding to the above execution example 3 is displayed in a distinguished manner. When the variable domain value distribution diagram in the different execution case on the display diagram is not the fifth generation, the variable domain value distribution diagram in the different execution case of the fifth generation is displayed in the same manner as in the description of the operation (part 3). After that, it is distinguished.
[0139]
As a result, it is possible to observe the variable domain value distribution state in different execution cases in a specific generation and the state of the best chromosome in the execution case.
(9) Description of Operation of the Present Invention (No. 9)
In this case, a generation change diagram of the best chromosome (FIG. 11) and a generation change diagram of the fitness deviation (FIG. 12) are used from the display diagram displayed on the display screen 1-0. For example, when the second generation on the generation coordinate scale is designated by the position specifying device 2 on the generation change diagram of the best chromosome, each execution is performed from the display diagram indication transmitting means on the ninth display unit (not shown) in the same manner as described above. A transmission signal is output which instructs to distinguish and display the second generation of the case.
[0140]
This is received by the display indicator indicating means of the tenth display unit (not shown) for creating the display diagram of FIG. 12, and all of the generation change diagrams of the fitness deviation on the display screen 1-0 belonging to the second generation are displayed. , That is, the second-generation mark figures of execution examples 1, 2, and 3 are distinguished and displayed.
[0141]
Conversely, when the position specifying device 2 designates, for example, the third generation on the generation coordinate scale on the generation variation diagram of the fitness deviation, all the belongings belonging to the third generation on the generation variation diagram of the best chromosome are processed in the same manner as described above. Are displayed in a divided manner.
[0142]
Thereby, the state of the generation change of the best chromosome and the generation change of the fitness deviation can be observed.
C, another embodiment of the present invention
Although the above description has been made with reference to the example of the chromosome of the bit string expression type, the present invention is not limited to this, and can be applied to the chromosome of the ordered expression type. A chromosome distribution diagram in the case of the ordered phenotype chromosome expression will be described with reference to FIG.
[0143]
FIG. 13 illustrates the traveling salesman problem for cities A to H as one example in the case of the ordered phenotype chromosome representation, and illustrates the distribution of chromosomes over a domain. As shown in FIG. 13 (A), chromosome 1 is represented by “EBADCGHF”, chromosome 2 is represented by “GEFBCADH”,... Chromosome 8 is represented by “GDCAEBFH”.
[0144]
FIG. 13B shows the adjacency between the two cities of chromosomes 1 to 8, in which the adjacent cities are indicated by black marks and the non-adjacent cities are indicated by white marks. And city B are adjacent to each other, so they are marked black, and cities A and C are not adjacent to each other, so they are marked white. Similarly, on chromosome 2, cities A and B are not adjacent to each other on the circuit, and thus are marked white, and cities A and C are adjacent and marked black. Therefore, it can be understood that cities AB, AD, BE, GH, etc., each having a large number of black marks in each of the chromosomes 1 to 8, have high adjacency.
[0145]
A method for creating a cluster dendrogram of a chromosome population as shown in FIG. 3 using such ordered phenotype chromosomes is described in Japanese Patent Application No. Hei 7-238389.
[0146]
Now, when FIG. 13 (B) and the cluster dendrogram created based on it are displayed on the display screen 1-0, the horizontal axis scale indicating the chromosome number of FIG. 13 (B) or FIG. 13 (B) Is designated by the position specifying device 2, the chromosome of the scale number is displayed in the case of the horizontal axis scale, and the number of the chromosome including this mark is displayed in the case of the mark. The display diagram of the display unit displaying the adjacency between the two cities is sent from the indicator transmission unit to the display diagram indicator unit of the display unit on the side of the cluster tree diagram, and the cluster tree diagram of the same generation is displayed. The corresponding chromosome is displayed.
[0147]
Conversely, when a chromosome on the cluster dendrogram is designated, all the corresponding chromosome marks or the chromosome numbers on the horizontal axis scale in the diagram showing the adjacency between the two cities are displayed in the same manner as described above.
[0148]
As a result, it is possible to observe the cluster dendrogram of the chromosome population and the state of the adjacency between the two cities of the ordered phenotype chromosome.
In the present invention, the ordinal phenotype chromosome expression, and the horizontal axis in FIG. 14 is the best chromosome of execution cases 1, 2, 3,... Can show the greatest chromosome adjacency. This can be used instead of FIG.
[0149]
In the above description, an example has been described in which one drawing is designated so as to be distinguished from the other drawing. However, the present invention is not limited to this. It is also possible to display the corresponding parts in.
[0150]
In the above description, an example was described in which the chromosome population for each generation and various figures created thereby were stored in the auxiliary storage device 3. However, the present invention is not limited to this, and the main storage device is not limited thereto. Can also be stored.
[0151]
Also, an example has been described in which data obtained by performing a genetic algorithm operation process as shown in FIGS. 15 and 16 by a genetic algorithm operation device (not shown) is used, but the present invention is of course not limited to this. Alternatively, data obtained by performing a genetic algorithm operation process by the control unit 5 may be used.
[0152]
According to the present invention, the execution result of the genetic algorithm can be observed while comparing a plurality of drawings, so that the judgment for predicting the execution result becomes accurate.
[0153]
【The invention's effect】
According to the present invention, the following excellent effects can be obtained.
(1) A plurality of types of display diagrams are displayed on the display screen, and when a part of a figure on a certain display figure is designated by the position designating means, the corresponding figure on another display figure is changed in color, pattern, or blinking. , Such as an arrow figure, such as a visual representation, such as a cluster tree diagram of a chromosome population and a variable domain distribution diagram of a chromosome population, such as the same chromosome on multiple types of display diagrams Since the chromosome-related information can be observed, it can help to understand the characteristics of the chromosome information, for example, the characteristics of the chromosome population distribution.
[0154]
(2) The generation change diagram of the fitness distribution of the chromosome population of a certain range of generations and the generation change diagram of another type of chromosome distribution can be displayed on the display screen, and the correspondence can be observed. It is possible to easily recognize the positional relationship of the same chromosome on a plurality of kinds of display diagrams, thereby helping to understand the characteristics of the chromosome population distribution.
[0155]
(3) A display diagram for a specific generation, such as a cluster tree diagram of a chromosome population of a specific generation, and a display diagram for a generation range, such as a generation change diagram of a chromosome population fitness distribution, are displayed on a display screen. Since the correspondence can be observed, the positional relationship of the same chromosome on a plurality of types of display diagrams can be easily recognized, which can help to understand the characteristics of the chromosome population distribution.
[0156]
(4) In the display screen, the generation change diagram of the maximum fitness and the generation change locus diagram of the fitness maximum value and the fitness deviation, the analysis amount in each generation of the chromosome group of the genetic algorithm in a plurality of generation ranges Since it is possible to display and observe the correspondence, it is easy to recognize the positional relationship of the analysis amount of the same generation on multiple types of display diagrams, and thereby, the characteristic of the generation change of the analysis amount of the chromosome population Can help you understand.
[0157]
(5) On the display screen, a plurality of types of display diagrams of a chromosome in a specific generation or a generation range such as a generation change diagram of a chromosome population fitness distribution and a generation change locus diagram of a fitness maximum value and a fitness deviation are shown. Since it is possible to display and observe the correspondence, it is possible to easily recognize the positional relationship of the analysis amount of the same generation on multiple types of display diagrams, and thereby to characterize the generation change of the analysis amount of the chromosome population. Can help you understand.
[0158]
(6) A plurality of execution examples of the genetic algorithm executed under different conditions are displayed on the display screen, such as a distribution map of fitness values in different execution cases and a distribution diagram of variable domain values in different execution cases. In addition, since the correspondence can be observed, it is easy to recognize the positional relationship of the analysis amount of the same execution case on a plurality of types of display diagrams. Can help you understand.
[0159]
(7) Display multiple types of display diagrams of execution examples for the generation range, such as the generation change diagram of the best chromosome and the generation change diagram of the fitness deviation, on the display screen, and observe the corresponding relationship. Can easily recognize the positional relationship between the amounts of analysis of the same execution case on multiple types of display diagrams, thereby helping to understand the characteristics of the difference in generation change of execution cases under different conditions. Can be.
[0160]
(8) On the display screen, a display diagram for a specific generation, such as a distribution diagram of variable domain values in different execution cases, and a display diagram for a generation range, such as a change diagram of the generation of the best chromosome, are displayed. Since the correspondence can be observed, it is easy to recognize the positional relationship between the amounts of analysis of the same execution case on multiple types of display diagrams, so that it is possible to understand the difference between execution cases under different conditions. I can help.
[0161]
(9) Display multiple types of display diagrams of execution examples for a certain generation range, such as a generation change diagram of the best chromosome and a generation change diagram of the fitness deviation, on the display screen, and observe the correspondence between them. Can easily recognize differences in the amount of analysis of execution cases in the same generation on multiple types of display diagrams, thereby helping to understand differences in execution cases under different conditions. .
[Brief description of the drawings]
FIG. 1 is a principle configuration diagram of the present invention.
FIG. 2 is an embodiment of the present invention.
FIG. 3 is a cluster dendrogram of a chromosome population.
FIG. 4 is a variable domain distribution map of a chromosome population.
FIG. 5 is a generation change diagram of a chromosome population fitness distribution.
FIG. 6 is a generation change diagram of a distribution of a value of a variable 1;
FIG. 7 is a generation change diagram of the maximum fitness.
FIG. 8 is a generation change locus diagram of a fitness maximum value and a fitness deviation.
FIG. 9 is a distribution diagram of fitness values in different execution cases.
FIG. 10 is a distribution diagram of variable domain values in different execution cases.
FIG. 11 is a generation change diagram of the best chromosome.
FIG. 12 is a generation change diagram of a fitness deviation.
FIG. 13 is a chromosome distribution diagram in the case of ordered phenotype chromosome expression.
FIG. 14 is a distribution map of best chromosomes in different execution cases.
FIG. 15 is an explanatory diagram of a conventional example of an order expression.
FIG. 16 is an explanatory diagram of a conventional example of bit string expression.
[Explanation of symbols]
1 Image display device
1-0 Display screen
2 Positioning device
3 Auxiliary storage device
4 Character input device
5 control part
6-1 First display section
6-2 Second display unit
6-n n-th display unit
7-1 Transmitter
7-2 Transmitter
8-1 Receiver
8-2 Receiver
9-1 Display diagram creation means
9-2 Display diagram creation means
10-1 Means for detecting a display figure indicator
10-2 Display Diagram Indicator Detecting Means
11-1 Means for indicating and indicating objects
11-2 Means for indicating indications and indicating objects
12-1 Display diagram control unit
12-2 Display diagram control unit
13-1 Storage Means
13-2 Storage means

Claims (7)

Image display means having a display screen for displaying the operation state of the genetic algorithm and chromosome information transmitted from the display unit, and genetic algorithm analysis display processing comprising position designation means for designating a part of a display diagram on the display screen In the device,
Provide multiple display units,
In each display,
Display diagram creating means for creating a display diagram to be displayed on the display screen,
A display diagram indicating object detecting means for detecting the indicating object specified by the position specifying means,
A display diagram for transmitting an indicator signal indicating the indicator, an indicator transmitting means,
A display diagram indicator receiving means for receiving an indicator signal transmitted from another display unit,
A display diagram indicating object indicating means for distinguishing and displaying a figure corresponding to the received display object from others is provided,
A genetic algorithm analysis, wherein when a figure on a certain display diagram is designated by the position designating means, a corresponding portion of the designated figure on another display figure is distinguished and displayed by the display figure designating means showing means. Display processing device. On the display screen, a generation change diagram of a certain distribution of a chromosome population of a certain range of generations and a generation change diagram of another type of distribution of chromosomes are displayed, and when the chromosome of one diagram is designated, the other diagram is displayed. 2. The genetic algorithm analysis display processing device according to claim 1, wherein a mark figure corresponding to the chromosome is distinguished and displayed. On the display screen, a display diagram for a specific generation and a display diagram for a generation range are displayed, and when one of the chromosomes in one diagram is designated, the mark figure corresponding to this chromosome in the other diagram is displayed in a distinguishing manner. The genetic algorithm analysis display processing device according to claim 1, wherein the processing is performed. On the display screen, the amount of analysis in each generation of the chromosome population of the genetic algorithm is displayed for a range from a certain generation to a certain generation, and when the generation information display part of one figure is designated by the position designation means, 2. The genetic algorithm analysis display processing device according to claim 1, wherein the corresponding generation information display portions on the display diagram are distinguished and displayed. On the display screen, a display diagram of a specific generation or a generation range for a chromosome and a generation change diagram of a distribution of a chromosome population of a certain range of generations are displayed, and the generation information display portion of one of the diagrams is displayed. 2. The genetic algorithm analysis display processing device according to claim 1, wherein, when the position is designated by the position designation means, a corresponding generation information display portion on another display diagram is distinguished and displayed. On the display screen, and executes a plurality of different conditions, the genetic algorithm, with respect to a plurality of execution cases, certain analysis amount with each execution case in particular generation is analyzed amount of the illustrating the different execution examples Display a distribution diagram and a distribution diagram of another analysis amount in a different execution case that illustrates another type of analysis amount of the same execution case, and display an execution case of a specific condition in one distribution diagram 2. The genetic algorithm analysis display processing device according to claim 1, wherein when a portion is designated by the position specifying means, a display portion of a corresponding execution case on another distribution diagram is displayed in a distinguished manner. On the display screen, a display diagram for a specific generation and a display diagram for a generation range are displayed, and when a display portion of an execution example of one diagram is designated by the position designating unit, a correspondence on another display diagram is displayed. 2. The genetic algorithm analysis display processing device according to claim 1, wherein the display part of the execution case to be executed is displayed in a distinguished manner.

Images