記事「母集団最適化アルゴリズム：微小人工免疫系(Micro-AIS)」についてのディスカッション

MetaQuotes 2024.05.13 17:40

新しい記事「母集団最適化アルゴリズム：微小人工免疫系(Micro-AIS)」はパブリッシュされました:

この記事では、身体の免疫系の原理に基づいた最適化手法、つまりAISを改良した微小人工免疫系（Micro Artificial Immune System：Micro-AIS）について考察します。Micro-AISは、より単純な免疫系のモデルと単純な免疫情報処理操作を用います。また、この記事では、従来のAISと比較した場合のMicro-AISの利点と欠点についても触れています。

AISアルゴリズムは、抗原（入力）、抗体（解）、キラー細胞（最適化プロセス）の概念を用いてこれらのプロセスをモデル化し、問題を最適に解決します。抗原は最適化されるべき入力を表します。抗体はこの問題に対する潜在的な解です。キラー細胞は、最適化問題に対する最適解を探索する最適化プロセスです。

人工免疫系(AIS)最適化法は1990年代に提案されました。この手法に関する初期の研究は1980年代半ばまでさかのぼり、Farmer、Packard、Perelson (1986)、Bersini and Varela (1990)が大きく貢献しています。

それ以来、AIS法は発展を続け、科学界で継続的な研究の対象となっています。この方法は、様々な最適化問題や学習問題への応用と同様に、多くのバリエーションや修正が提案されています。身体の免疫系は、感染症や腫瘍などの外的影響から身を守るためにも重要な役割を果たしています。敵対的なエージェントを識別し、将来の使用のために情報を保存する能力を維持しながら、異常を認識し、検出し、攻撃する能力を持ちます。

Micro-AIS (Micro-Immune Algorithm)は、最適化問題を解くために開発された免疫系(AIS)アルゴリズムを改良したものです。従来のAISとの違いは、より単純な免疫系のモデルと、より単純な免疫情報処理操作を用いる点です。

作者: Andrey Dik

fxsaber 2024.01.16 11:29 #1

№	AO	Description	Hilly			Hilly final	Forest			Forest final	Megacity (discrete)			Megacity final	Final result	% of MAX
№	AO	Description	10 p (5 F)	50 p (25 F)	1000 p (500 F)	Hilly final	10 p (5 F)	50 p (25 F)	1000 p (500 F)	Forest final	10 p (5 F)	50 p (25 F)	1000 p (500 F)	Megacity final	Final result	% of MAX
1	(P+O)ES	(P+O) evolution strategies	0,99934	0,91895	0,56297	2,48127	1,00000	0,93522	0,39179	2,32701	0,83167	0,64433	0,21155	1,68755	6,496	72,18
2	SDSm	stochastic diffusion search M	0,93066	0,85445	0,39476	2,17988	0,99983	0,89244	0,19619	2,08846	0,72333	0,61100	0,10670	1,44103	5,709	63,44

この表の関連記事へのリンクをお願いします。

Inquiring 2024.01.17 10:15 #2

ニューラルネットワーク、アルゴリズム、最適化、学習、どれも素晴らしい。

メッセージを読み取るのにどんなアルゴリズムが使えるだろうか？

システム分析において重要なのは、システムを構成する要素とその間の機能的関係であり、システムが存在する意味（目的）によって結びついている。

機能分析を適用すべきところで、確率分析を適用することがどれほど正当なことか。

問題は横道にそれながらも、プロセスの根本的な基礎を理解する方向にあるのかもしれない。

fxsaber 2024.01.19 18:18 #3

取引、自動取引システム、取引戦略のテストに関するフォーラム

ライブラリ: Input_Struct

fxsaber, 2024.01.19 18:16

カスタム最適化アルゴリズム用ライブラリの使用例。

fxsaber 2024.01.19 22:30 #4

種の関数を最適化するのは難しい？

double FF( const double &Arg[] )
{
  double Res = 1;

  for (uint i = ArraySize(Arg); (bool)i--;)
    Res *= MathSin(Arg[i]);

  return(Res);
}

最良の値は1に近いはずだ。しかし、この連載記事のアルゴリズムはこの最大値に近づけることができるのだろうか？

Andrey Dik 2024.01.20 07:40 #5

fxsaber #:
種の関数を最適化するのは難しい？最良の値は1に近いはずだ。しかし、この連載記事のアルゴリズムはこの最大値に近づけることができるだろうか？

私の分類では、この関数は単純なものに属する。

Andrey Dik 2024.01.20 07:58 #6

fxsaber #:

この表の関連記事へのリンクをお願いします

次の記事からリンクがあります。比較的簡単な方法を見つけました。

Stanislav Korotky 2024.01.20 15:52 #7

fxsaber #:

このシリーズのアルゴリズムを長い間オプティマイザーに引きずり込もうとした（並列化したかった）のだが、大失敗してしまった - https://www.mql5.com/en/forum/454524/page2#comment_50233782。

MT5 forward testing bugs.

2023.10.20
www.mql5.com

This post is about a some weirdly related (I think) bugs that happen when forward testing optimized results. Please read through it all...

fxsaber 2024.01.20 15:57 #8

Stanislav Korotky #:

入れた。

Optimization - несколько алгоритмов оптимизации в одном месте.

www.mql5.com

Получилось собрать в одном месте сразу несколько алгоритмов оптимизации и создать простой механизм их использования. Механизм. Помещаем советник в Тестер и используем GUI вкладки Inputs , чтобы

fxsaber 2024.01.20 17:55 #9

取引、自動取引システム、取引戦略のテストに関するフォーラム

論文「集団最適化アルゴリズム：マイクロ人工免疫システム（Micro-AIS）アルゴリズム」についての議論。

fxsaber, 2024.01.19 22:30

種の関数を最適化するのは難しいですか？

double FF( const double &Arg[] )
{
  double Res = 1;

  for (uint i = ArraySize(Arg); (bool)i--;)
    Res *= MathSin(Arg[i]);

  return(Res);
}

最適な値は1に近いはずです。しかし、この連載記事のアルゴリズムはこの最大値に近づけることができるのだろうか？

テスト済み。

#define  dInput01 X1
#define  dInput02 X2
#define  dInput03 X3
#define  dInput04 X4
#define  dInput05 X5
#define  dInput06 X6
#define  dInput07 X7

#include <fxsaber\Input_Struct\Input_Struct.mqh> //https://www.mql5.com/ja/code/47932

INPUT_STRUCT inInputs;

MACROS_INPUT(double, X1, 0);
MACROS_INPUT(double, X2, 0);
MACROS_INPUT(double, X3, 0);
MACROS_INPUT(double, X4, 0);
MACROS_INPUT(double, X5, 0);
MACROS_INPUT(double, X6, 0);
MACROS_INPUT(double, X7, 0);

// 社内オプティマイザーのFFはその一例である。
double OnTester()
{
  return(MathSin(inInputs.X1) *
         MathSin(inInputs.X2) *
         MathSin(inInputs.X3) *
         MathSin(inInputs.X4) *
         MathSin(inInputs.X5) *
         MathSin(inInputs.X6) *
         MathSin(inInputs.X7));
}

#include <fxsaber\Optimization\Optimization_Addon.mqh> //https://www.mql5.com/ru/blogs/post/755815

入力。

カスタム・オプティマイザー。

カスタム。

PSO Finished 15835 of 35000 planned passes: true
BestResult = 0.9884554736115849: X1 = 99.0, X2 = 99.0, X3 = 11.0, X4 = 77.0, X5 = 14.0, X6 = 11.0, X7 = 33.0
Check = 0.9884554736115849: X1 = 99.0, X2 = 99.0, X3 = 11.0, X4 = 77.0, X5 = 14.0, X6 = 11.0, X7 = 33.0

01: OPTIMIZATION_METHOD_AO_Micro_AIS
OPTIMIZATION_METHOD_AO_Micro_AIS
BestResult = 0.6914924547679845: X1 = 17.0, X2 = 89.0, X3 = 61.0, X4 = 33.0, X5 = 71.0, X6 = 64.0, X7 = 8.0
Check = 0.6914924547679845: X1 = 17.0, X2 = 89.0, X3 = 61.0, X4 = 33.0, X5 = 71.0, X6 = 64.0, X7 = 8.0

02: OPTIMIZATION_METHOD_AO_POES
OPTIMIZATION_METHOD_AO_POES
BestResult = 0.9268682527605293: X1 = 55.0, X2 = 80.0, X3 = 27.0, X4 = 99.0, X5 = 8.0, X6 = 52.0, X7 = 11.0
Check = 0.9268682527605293: X1 = 55.0, X2 = 80.0, X3 = 27.0, X4 = 99.0, X5 = 8.0, X6 = 52.0, X7 = 11.0

03: OPTIMIZATION_METHOD_AO_P_O_ES
OPTIMIZATION_METHOD_AO_P_O_ES
BestResult = 0.7717845794829589: X1 = 11.0, X2 = 49.0, X3 = 74.0, X4 = 30.0, X5 = 11.0, X6 = 77.0, X7 = 43.0
Check = 0.7717845794829589: X1 = 11.0, X2 = 49.0, X3 = 74.0, X4 = 30.0, X5 = 11.0, X6 = 77.0, X7 = 43.0

04: OPTIMIZATION_METHOD_AO_SC
OPTIMIZATION_METHOD_AO_SC
BestResult = 0.5703083565001157: X1 = 4.0, X2 = 39.0, X3 = 20.0, X4 = 93.0, X5 = 8.0, X6 = 20.0, X7 = 33.0
Check = 0.5703083565001157: X1 = 4.0, X2 = 39.0, X3 = 20.0, X4 = 93.0, X5 = 8.0, X6 = 20.0, X7 = 33.0

05: OPTIMIZATION_METHOD_AO_SIA
OPTIMIZATION_METHOD_AO_SIA
BestResult = 0.3770511069126069: X1 = 30.0, X2 = 55.0, X3 = 49.0, X4 = 77.0, X5 = 100.0, X6 = 65.0, X7 = 27.0
Check = 0.3770511069126069: X1 = 30.0, X2 = 55.0, X3 = 49.0, X4 = 77.0, X5 = 100.0, X6 = 65.0, X7 = 27.0

06: OPTIMIZATION_METHOD_AO_SA
OPTIMIZATION_METHOD_AO_SA
BestResult = 0.4195625904721657: X1 = 58.0, X2 = 77.0, X3 = 27.0, X4 = 40.0, X5 = 70.0, X6 = 14.0, X7 = 70.0
Check = 0.4195625904721657: X1 = 58.0, X2 = 77.0, X3 = 27.0, X4 = 40.0, X5 = 70.0, X6 = 14.0, X7 = 70.0

07: OPTIMIZATION_METHOD_AO_NMm
OPTIMIZATION_METHOD_AO_NMm
BestResult = 0.8314291991406518: X1 = 30.0, X2 = 46.0, X3 = 99.0, X4 = 11.0, X5 = 96.0, X6 = 39.0, X7 = 74.0
Check = 0.8314291991406518: X1 = 30.0, X2 = 46.0, X3 = 99.0, X4 = 11.0, X5 = 96.0, X6 = 39.0, X7 = 74.0

08: OPTIMIZATION_METHOD_AO_DE
OPTIMIZATION_METHOD_AO_DE
BestResult = 0.514763435265798: X1 = 33.0, X2 = 39.0, X3 = 49.0, X4 = 20.0, X5 = 73.0, X6 = 20.0, X7 = 58.0
Check = 0.514763435265798: X1 = 33.0, X2 = 39.0, X3 = 49.0, X4 = 20.0, X5 = 73.0, X6 = 20.0, X7 = 58.0

09: OPTIMIZATION_METHOD_AO_SDOm
OPTIMIZATION_METHOD_AO_SDOm
BestResult = 0.6248310950237546: X1 = 55.0, X2 = 61.0, X3 = 20.0, X4 = 71.0, X5 = 26.0, X6 = 74.0, X7 = 36.0
Check = 0.6248310950237546: X1 = 55.0, X2 = 61.0, X3 = 20.0, X4 = 71.0, X5 = 26.0, X6 = 74.0, X7 = 36.0

10: OPTIMIZATION_METHOD_AO_IWDm
OPTIMIZATION_METHOD_AO_IWDm
BestResult = 0.6582185170915256: X1 = 33.0, X2 = 24.0, X3 = 61.0, X4 = 55.0, X5 = 46.0, X6 = 36.0, X7 = 1.0
Check = 0.6582185170915256: X1 = 33.0, X2 = 24.0, X3 = 61.0, X4 = 55.0, X5 = 46.0, X6 = 36.0, X7 = 1.0

11: OPTIMIZATION_METHOD_AO_CSS
OPTIMIZATION_METHOD_AO_CSS
BestResult = 0.17125241139972677: X1 = 11.0, X2 = 5.0, X3 = 11.0, X4 = 37.0, X5 = 56.0, X6 = 65.0, X7 = 37.0
Check = 0.17125241139972677: X1 = 11.0, X2 = 5.0, X3 = 11.0, X4 = 37.0, X5 = 56.0, X6 = 65.0, X7 = 37.0

12: OPTIMIZATION_METHOD_AO_SDS
OPTIMIZATION_METHOD_AO_SDS
BestResult = 0.7015125972513457: X1 = 17.0, X2 = 46.0, X3 = 27.0, X4 = 39.0, X5 = 77.0, X6 = 71.0, X7 = 86.0
Check = 0.7015125972513457: X1 = 17.0, X2 = 46.0, X3 = 27.0, X4 = 39.0, X5 = 77.0, X6 = 71.0, X7 = 86.0

13: OPTIMIZATION_METHOD_AO_SDSm
OPTIMIZATION_METHOD_AO_SDSm
BestResult = 0.8318883232825393: X1 = 77.0, X2 = 14.0, X3 = 14.0, X4 = 30.0, X5 = 80.0, X6 = 49.0, X7 = 24.0
Check = 0.8318883232825393: X1 = 77.0, X2 = 14.0, X3 = 14.0, X4 = 30.0, X5 = 80.0, X6 = 49.0, X7 = 24.0

14: OPTIMIZATION_METHOD_AO_MEC
OPTIMIZATION_METHOD_AO_MEC
BestResult = 0.821421124921697: X1 = 99.0, X2 = 58.0, X3 = 90.0, X4 = 27.0, X5 = 14.0, X6 = 80.0, X7 = 96.0
Check = 0.821421124921697: X1 = 99.0, X2 = 58.0, X3 = 90.0, X4 = 27.0, X5 = 14.0, X6 = 80.0, X7 = 96.0

15: OPTIMIZATION_METHOD_AO_SFL
OPTIMIZATION_METHOD_AO_SFL
BestResult = 0.7123520662251704: X1 = 49.0, X2 = 52.0, X3 = 80.0, X4 = 93.0, X5 = 52.0, X6 = 87.0, X7 = 14.0
Check = 0.7123520662251704: X1 = 49.0, X2 = 52.0, X3 = 80.0, X4 = 93.0, X5 = 52.0, X6 = 87.0, X7 = 14.0

16: OPTIMIZATION_METHOD_AO_EM
OPTIMIZATION_METHOD_AO_EM
BestResult = 0.4739892519704631: X1 = 39.0, X2 = 96.0, X3 = 49.0, X4 = 54.0, X5 = 93.0, X6 = 8.0, X7 = 11.0
Check = 0.4739892519704631: X1 = 39.0, X2 = 96.0, X3 = 49.0, X4 = 54.0, X5 = 93.0, X6 = 8.0, X7 = 11.0

17: OPTIMIZATION_METHOD_AO_SSG
OPTIMIZATION_METHOD_AO_SSG
BestResult = 0.7570642423726676: X1 = 5.0, X2 = 49.0, X3 = 30.0, X4 = 96.0, X5 = 14.0, X6 = 55.0, X7 = 89.0
Check = 0.7570642423726676: X1 = 5.0, X2 = 49.0, X3 = 30.0, X4 = 96.0, X5 = 14.0, X6 = 55.0, X7 = 89.0

18: OPTIMIZATION_METHOD_AO_MA
OPTIMIZATION_METHOD_AO_MA
BestResult = 0.7831093525101701: X1 = 93.0, X2 = 36.0, X3 = 17.0, X4 = 58.0, X5 = 42.0, X6 = 61.0, X7 = 74.0
Check = 0.7831093525101701: X1 = 93.0, X2 = 36.0, X3 = 17.0, X4 = 58.0, X5 = 42.0, X6 = 61.0, X7 = 74.0

19: OPTIMIZATION_METHOD_AO_HS
OPTIMIZATION_METHOD_AO_HS

Error optimization!

20: OPTIMIZATION_METHOD_AO_GSA
OPTIMIZATION_METHOD_AO_GSA
BestResult = 0.020184193323560605: X1 = 9.0, X2 = 27.0, X3 = 54.0, X4 = 6.0, X5 = 77.0, X6 = 75.0, X7 = 23.0
Check = 0.020184193323560605: X1 = 9.0, X2 = 27.0, X3 = 54.0, X4 = 6.0, X5 = 77.0, X6 = 75.0, X7 = 23.0

21: OPTIMIZATION_METHOD_AO_GSA_Stars
OPTIMIZATION_METHOD_AO_GSA_Stars

Error optimization!

22: OPTIMIZATION_METHOD_AO_BFO
OPTIMIZATION_METHOD_AO_BFO
BestResult = 0.7322059190279094: X1 = 20.0, X2 = 11.0, X3 = 52.0, X4 = 49.0, X5 = 89.0, X6 = 36.0, X7 = 99.0
Check = 0.7322059190279094: X1 = 20.0, X2 = 11.0, X3 = 52.0, X4 = 49.0, X5 = 89.0, X6 = 36.0, X7 = 99.0

23: OPTIMIZATION_METHOD_AO_IWO
OPTIMIZATION_METHOD_AO_IWO
BestResult = 0.7392111937754324: X1 = 20.0, X2 = 11.0, X3 = 52.0, X4 = 49.0, X5 = 89.0, X6 = 37.0, X7 = 100.0
Check = 0.24076952243473274: X1 = 20.0, X2 = 11.0, X3 = 52.0, X4 = 49.0, X5 = 89.0, X6 = 37.0, X7 = 100.0

24: OPTIMIZATION_METHOD_AO_BA
OPTIMIZATION_METHOD_AO_BA
BestResult = 0.35033516894855804: X1 = 98.0, X2 = 49.0, X3 = 92.0, X4 = 77.0, X5 = 96.0, X6 = 99.0, X7 = 21.0
Check = 0.35033516894855804: X1 = 98.0, X2 = 49.0, X3 = 92.0, X4 = 77.0, X5 = 96.0, X6 = 99.0, X7 = 21.0

25: OPTIMIZATION_METHOD_AO_FAm
OPTIMIZATION_METHOD_AO_FAm
BestResult = 0.8628261244286874: X1 = 61.0, X2 = 33.0, X3 = 93.0, X4 = 55.0, X5 = 30.0, X6 = 49.0, X7 = 55.0
Check = 0.8628261244286874: X1 = 61.0, X2 = 33.0, X3 = 93.0, X4 = 55.0, X5 = 30.0, X6 = 49.0, X7 = 55.0

26: OPTIMIZATION_METHOD_AO_FSS
OPTIMIZATION_METHOD_AO_FSS
BestResult = 0.6586267117021989: X1 = 90.0, X2 = 17.0, X3 = 30.0, X4 = 11.0, X5 = 11.0, X6 = 89.0, X7 = 46.0
Check = 0.6586267117021989: X1 = 90.0, X2 = 17.0, X3 = 30.0, X4 = 11.0, X5 = 11.0, X6 = 89.0, X7 = 46.0

27: OPTIMIZATION_METHOD_AO_COAm
OPTIMIZATION_METHOD_AO_COAm
BestResult = 0.751387775021677: X1 = 33.0, X2 = 74.0, X3 = 89.0, X4 = 52.0, X5 = 2.0, X6 = 8.0, X7 = 99.0
Check = 0.751387775021677: X1 = 33.0, X2 = 74.0, X3 = 89.0, X4 = 52.0, X5 = 2.0, X6 = 8.0, X7 = 99.0

28: OPTIMIZATION_METHOD_AO_GWO
OPTIMIZATION_METHOD_AO_GWO
BestResult = 0.7905125996746682: X1 = 64.0, X2 = 24.0, X3 = 58.0, X4 = 11.0, X5 = 39.0, X6 = 36.0, X7 = 55.0
Check = 0.7905125996746682: X1 = 64.0, X2 = 24.0, X3 = 58.0, X4 = 11.0, X5 = 39.0, X6 = 36.0, X7 = 55.0

29: OPTIMIZATION_METHOD_AO_ABC
OPTIMIZATION_METHOD_AO_ABC
BestResult = 0.2279828722733523: X1 = 37.0, X2 = 49.0, X3 = 45.0, X4 = 96.0, X5 = 86.0, X6 = 54.0, X7 = 89.0
Check = 0.2279828722733523: X1 = 37.0, X2 = 49.0, X3 = 45.0, X4 = 96.0, X5 = 86.0, X6 = 54.0, X7 = 89.0

30: OPTIMIZATION_METHOD_AO_ACOm
OPTIMIZATION_METHOD_AO_ACOm
BestResult = 0.7283588705105443: X1 = 58.0, X2 = 36.0, X3 = 46.0, X4 = 58.0, X5 = 77.0, X6 = 42.0, X7 = 46.0
Check = 0.7283588705105443: X1 = 58.0, X2 = 36.0, X3 = 46.0, X4 = 58.0, X5 = 77.0, X6 = 42.0, X7 = 46.0

31: OPTIMIZATION_METHOD_AO_PSO
OPTIMIZATION_METHOD_AO_PSO
BestResult = 0.5892210470192797: X1 = 52.0, X2 = 52.0, X3 = 68.0, X4 = 62.0, X5 = 86.0, X6 = 77.0, X7 = 30.0
Check = 0.5892210470192797: X1 = 52.0, X2 = 52.0, X3 = 68.0, X4 = 62.0, X5 = 86.0, X6 = 77.0, X7 = 30.0

32: OPTIMIZATION_METHOD_AO_RND
OPTIMIZATION_METHOD_AO_RND
BestResult = 0.6663782757838177: X1 = 4.896755719304697, X2 = 61.0, X3 = 23.0, X4 = 8.0, X5 = 52.0, X6 = 67.0, X7 = 58.0
Check = 0.6663782757838177: X1 = 4.896755719304697, X2 = 61.0, X3 = 23.0, X4 = 8.0, X5 = 52.0, X6 = 67.0, X7 = 58.0

ZY 何かがIWOで正しく動作していない。たぶん、私が間違って移植したのだろう。

Vladimir Suslov 2024.01.20 19:05 #10

fxsaber #:

そして、なぜ最適化によってこの関数のパラメーターを探さなければならないのか？

答えは明白だ。

1 2 3 4 5

新しいコメント