{"id":251,"date":"2021-06-08T20:10:29","date_gmt":"2021-06-08T12:10:29","guid":{"rendered":"https:\/\/swordofmorning.com\/?p=251"},"modified":"2025-10-09T13:56:05","modified_gmt":"2025-10-09T05:56:05","slug":"matlab-tutorial-03-official-svm-and-nnw","status":"publish","type":"post","link":"https:\/\/swordofmorning.com\/index.php\/2021\/06\/08\/matlab-tutorial-03-official-svm-and-nnw\/","title":{"rendered":"Matlab Tutorial 03 official SVM and NNW"},"content":{"rendered":"

<\/div>
\n  \u672c\u6587\u4e3b\u8981\u4ecb\u7ecd\u5728Matlab\u4e2d\u4f7f\u7528\u5b98\u65b9\u5e93\u51fd\u6570\u63d0\u4f9b\u7684SVM\u652f\u6301\u5411\u91cf\u673a\u548cNNW\u795e\u7ecf\u7f51\u7edc\uff0c\u540c\u65f6\u5c06\u8fd9\u4e24\u79cd\u6a21\u578b\u5e94\u7528\u5230\u591a\u76ee\u6807\u5206\u7c7b\uff08Multi-Target Classfication\uff09\u4e0a\u3002<\/p>\n

\u4e00\u3001\u795e\u7ecf\u7f51\u7edc<\/h2>\n

1.1 newff()\u51fd\u6570\u53c2\u6570\u8bf4\u660e<\/h3>\n

  \u4f7f\u7528\u7684\u51fd\u6570newff()\u662f\u524d\u9988\u53cd\u5411\u4f20\u64ad\u7f51\u7edc\uff0c\u4e0b\u9762\u662f\u5b83\u7684\u5b9a\u4e49\u5f0f\uff1a<\/p>\n

net = newff(P,T,S)\nnet = newff(P,T,S,TF,BTF,BLF,PF,IPF,OPF,DDF)<\/code><\/pre>\n
    \n
  •   P\uff1a\u8f93\u5165\u53c2\u6570\u77e9\u9635\u3002(RxQ1)\uff0c\u5176\u4e2dQ1\u4ee3\u8868R\u5143\u7684\u8f93\u5165\u5411\u91cf\u3002\u5176\u6570\u636e\u610f\u4e49\u662f\u77e9\u9635P\u6709Q1\u5217\uff0c\u6bcf\u4e00\u5217\u90fd\u662f\u4e00\u4e2a\u6837\u672c\uff0c\u800c\u6bcf\u4e2a\u6837\u672c\u6709R\u4e2a\u5c5e\u6027\uff08\u7279\u5f81\uff09\u3002\u4e00\u822c\u77e9\u9635P\u9700\u8981\u5f52\u4e00\u5316\uff0c\u5373P\u7684\u6bcf\u4e00\u884c\u90fd\u5f52\u4e00\u5316\u5230[0 1]\u6216\u8005[-1 1]\u3002m * n\uff0cm\u662f\u7279\u5f81\u6570, n\u662f\u6837\u672c\u6570\u3002<\/li>\n
  •   T\uff1a\u76ee\u6807\u53c2\u6570\u77e9\u9635\u3002(SNxQ2)\uff0cQ2\u4ee3\u8868SN\u5143\u7684\u76ee\u6807\u5411\u91cf\u3002<\/li>\n
  •   S\uff1aN-1\u4e2a\u9690\u542b\u5c42\u7684\u6570\u76ee\uff08S\uff08i\uff09\u5230S\uff08N-1\uff09\uff09\uff0c\u9ed8\u8ba4\u4e3a\u7a7a\u77e9\u9635[]\u3002\u8f93\u51fa\u5c42\u7684\u5355\u5143\u6570\u76eeSN\u53d6\u51b3\u4e8eT\u3002\u8fd4\u56deN\u5c42\u7684\u524d\u9988BP\u795e\u7ecf\u7f51\u7edc<\/li>\n
  •   TF\uff1a\u76f8\u5173\u5c42\u7684\u4f20\u9012\u51fd\u6570\uff0c\u9ed8\u8ba4\u9690\u542b\u5c42\u4e3atansig\u51fd\u6570\uff0c\u8f93\u51fa\u5c42\u4e3apurelin\u51fd\u6570\u3002<\/li>\n
  •   BTF\uff1aBP\u795e\u7ecf\u7f51\u7edc\u5b66\u4e60\u8bad\u7ec3\u51fd\u6570\uff0c\u9ed8\u8ba4\u503c\u4e3atrainlm\u51fd\u6570\u3002<\/li>\n
  •   BLF\uff1a\u6743\u91cd\u5b66\u4e60\u51fd\u6570\uff0c\u9ed8\u8ba4\u503c\u4e3alearngdm\u3002<\/li>\n
  •   PF\uff1a\u6027\u80fd\u51fd\u6570\uff0c\u9ed8\u8ba4\u503c\u4e3amse\uff0c\u53ef\u9009\u62e9\u7684\u8fd8\u6709sse\uff0csae\uff0cmae\uff0ccrossentropy\u3002<\/li>\n
  •   IPF\uff0cOPF\uff0cDDF\u5747\u4e3a\u9ed8\u8ba4\u503c\u5373\u53ef\u3002<\/li>\n<\/ul>\n
    1.2 \u4f20\u9012\u51fd\u6570TF<\/h3>\n
      \n
    •   purelin\uff1a \u7ebf\u6027\u4f20\u9012\u51fd\u6570\u3002<\/li>\n
    •   tansig \uff1a\u6b63\u5207S\u578b\u4f20\u9012\u51fd\u6570\u3002<\/li>\n
    •   logsig \uff1a\u5bf9\u6570S\u578b\u4f20\u9012\u51fd\u6570\u3002\u3000<\/li>\n
    •   \u9690\u542b\u5c42\u548c\u8f93\u51fa\u5c42\u51fd\u6570\u7684\u9009\u62e9\u5bf9BP\u795e\u7ecf\u7f51\u7edc\u9884\u6d4b\u7cbe\u5ea6\u6709\u8f83\u5927\u5f71\u54cd\uff0c\u4e00\u822c\u9690\u542b\u5c42\u8282\u70b9\u8f6c\u79fb\u51fd\u6570\u9009\u7528 tansig\u51fd\u6570\u6216logsig\u51fd\u6570\uff0c\u8f93\u51fa\u5c42\u8282\u70b9\u8f6c\u79fb\u51fd\u6570\u9009\u7528tansig\u51fd\u6570\u6216purelin\u51fd\u6570\u3002<\/li>\n<\/ul>\n
      1.3 \u5b66\u4e60\u8bad\u7ec3\u51fd\u6570BTF<\/h3>\n
        \n
      •   traingd\uff1a\u6700\u901f\u4e0b\u964dBP\u7b97\u6cd5\u3002<\/li>\n
      •   traingdm\uff1a\u52a8\u91cfBP\u7b97\u6cd5\u3002<\/li>\n
      •   trainda\uff1a\u5b66\u4e60\u7387\u53ef\u53d8\u7684\u6700\u901f\u4e0b\u964dBP\u7b97\u6cd5\u3002<\/li>\n
      •   traindx\uff1a\u5b66\u4e60\u7387\u53ef\u53d8\u7684\u52a8\u91cfBP\u7b97\u6cd5\u3002<\/li>\n
      •   trainrp\uff1a\u5f39\u6027\u7b97\u6cd5\u3002
        \n\u3000\u3000<\/li>\n<\/ul>\n
          \u53d8\u68af\u5ea6\u7b97\u6cd5\uff1a<\/p>\n
          \n
        •   traincgf\uff08Fletcher-Reeves\u4fee\u6b63\u7b97\u6cd5\uff09<\/li>\n
        •   traincgp\uff08Polak_Ribiere\u4fee\u6b63\u7b97\u6cd5\uff09<\/li>\n
        •   traincgb\uff08Powell-Beale\u590d\u4f4d\u7b97\u6cd5\uff09<\/li>\n
        •   trainbfg\uff08BFGS \u62df\u725b\u987f\u7b97\u6cd5\uff09<\/li>\n
        •   trainoss\uff08OSS\u7b97\u6cd5\uff09<\/li>\n<\/ul>\n
          1.4 \u53c2\u6570\u8bf4\u660e<\/h3>\n
            \u901a\u8fc7net.trainParam\u53ef\u4ee5\u67e5\u770b\u53c2\u6570<\/p>\n
            \n
          •   Show Training Window Feedback showWindow: true<\/li>\n
          •   Show Command Line Feedback showCommandLine: false<\/li>\n
          •   Command Line Frequency show: \u4e24\u6b21\u663e\u793a\u4e4b\u95f4\u7684\u8bad\u7ec3\u6b21\u6570<\/li>\n
          •   Maximum Epochs epochs: \u8bad\u7ec3\u6b21\u6570<\/li>\n
          •   Maximum Training Time time: \u6700\u957f\u8bad\u7ec3\u65f6\u95f4\uff08\u79d2\uff09<\/li>\n
          •   Performance Goal goal: \u7f51\u7edc\u6027\u80fd\u76ee\u6807<\/li>\n
          •   Minimum Gradient min_grad: \u6027\u80fd\u51fd\u6570\u6700\u5c0f\u68af\u5ea6<\/li>\n
          •   Maximum Validation Checks max_fail: \u6700\u5927\u9a8c\u8bc1\u5931\u8d25\u6b21\u6570<\/li>\n
          •   Learning Rate lr: \u5b66\u4e60\u901f\u7387<\/li>\n
          •   Learning Rate Increase lr_inc: \u5b66\u4e60\u901f\u7387\u589e\u957f\u503c<\/li>\n
          •   Learning Rate lr_dec: \u5b66\u4e60\u901f\u7387\u4e0b\u964d\u503c<\/li>\n
          •   Maximum Performance Increase max_perf_inc:<\/li>\n
          •   Momentum Constant mc: \u52a8\u91cf\u56e0\u5b50<\/li>\n<\/ul>\n
            1.5 \u5b9e\u4f8b<\/h3>\n
            % \u521b\u5efa\u7f51\u7edc\nnet = newff(TriX, TriY, 9);\n\n% \u8bbe\u7f6e\u53c2\u6570\nnet.trainParam.epochs = 1000;\nnet.trainParam.goal = 1e-3;\nnet.trainParam.lr = 0.01;\n\n% \u8bad\u7ec3\nnet = train(net, TriX, TriY);\n\n% \u6d4b\u8bd5\nPreY = sim(net, TestX);<\/code><\/pre>\n
              \u503c\u5f97\u6ce8\u610f\u7684\u5730\u65b9\u662f\uff0c\u8fd9\u91cc\u7684X\u548cY\u9700\u8981\u662fm * n\uff0c\u5373m\u4e2a\u7279\u5f81* n\u4e2a\u6837\u672c\u6570\u3002<\/p>\n
            \u4e8c\u3001\u652f\u6301\u5411\u91cf\u673a<\/h2>\n
            2.1 fitcsvm()\u51fd\u6570\u53c2\u6570\u8bf4\u660e<\/h3>\n
              fitcsvm()\u672c\u8eab\u5e94\u7528\u4e8e\u5355\u76ee\u6807\u56de\u5f52\uff0c\u6211\u4eec\u53ef\u4ee5\u5c06\u591a\u76ee\u6807\u62c6\u5206\u6210\u591a\u4e2a\u5355\u76ee\u6807\u56de\u5f52\u4ee5\u5e94\u7528\u5230\u591a\u76ee\u6807\u4e0a\u3002<\/p>\n
            model = fitcsvm(X,Y,'ClassNames',{'negClass','posClass'},'Standardize',true,...\n        'KernelFunction','rbf','BoxConstraint',1);<\/code><\/pre>\n
              \n
            •   X\u662f\u8bad\u7ec3\u6837\u672c\uff0cnxm\u7684\u77e9\u9635\uff0cn\u662f\u6837\u672c\u6570\uff0cm\u662f\u7279\u5f81\u7ef4\u6570\u3002<\/li>\n
            •   Y\u662f\u6837\u672c\u6807\u7b7e\uff0cnx1\u7684\u77e9\u9635\uff0cn\u662f\u6837\u672c\u6570\u3002<\/li>\n
            •   \u2018ClassNames\u2019,{\u2018negClass\u2019,\u2018posClass\u2019} \u4e3a\u952e\u503c\u5bf9\u53c2\u6570\uff0c\u6307\u5b9a\u6b63\u8d1f\u7c7b\u522b\uff0c\u8d1f\u7c7b\u540d\u5728\u524d\uff0c\u6b63\u7c7b\u540d\u5728\u540e\uff0c\u4e0e\u6837\u672c\u6807\u7b7eY\u4e2d\u7684\u5143\u7d20\u5bf9\u5e94\u3002<\/li>\n
            •   \u2018Standardize\u2019,true \u4e3a\u952e\u503c\u5bf9\u53c2\u6570\uff0c\u6307\u793a\u8f6f\u4ef6\u662f\u5426\u5e94\u5728\u8bad\u7ec3\u5206\u7c7b\u5668\u4e4b\u524d\u4f7f\u9884\u6d4b\u671f\u6807\u51c6\u5316\u3002<\/li>\n
            •   \u2018KernelFunction\u2019,\u2018rbf\u2019 \u4e3a\u952e\u503c\u5bf9\u53c2\u6570\uff0c\u67093\u79cd \u2018linear\u2019\uff08\u9ed8\u8ba4\uff09, \u2018gaussian\u2019 (or \u2018rbf\u2019), \u2018polynomial\u2019\u3002<\/li>\n
            •   \u2018BoxConstraint\u2019,1 \u4e3a\u952e\u503c\u5bf9\u53c2\u6570\uff0c\u76f4\u89c2\u4e0a\u53ef\u4ee5\u7406\u89e3\u4e3a\u4e00\u4e2a\u60e9\u7f5a\u56e0\u5b50\uff08\u6216\u8005\u8bf4\u6b63\u5219\u53c2\u6570\uff09\uff0c\u8fd9\u4e2a\u53c2\u6570\u548csvmtrain\u91cc\u7684-c\u662f\u4e00\u4e2a\u9053\u7406\u3002\u5176\u5b9e\u9645\u4e0a\u6d89\u53ca\u5230\u8f6f\u95f4\u9694SVM\u7684\u95f4\u9694\uff08Margin\uff09\u5927\u5c0f\u3002<\/li>\n<\/ul>\n
              2.2 \u5b9e\u4f8b<\/h3>\n
               model1 = fitcsvm(TriX, TriY(:, 1));\n model2 = fitcsvm(TriX, TriY(:, 2));\n model3 = fitcsvm(TriX, TriY(:, 3));\n\n [label1, score1] = predict(model1, TestX);\n [label2, score2] = predict(model2, TestX);\n [label3, score3] = predict(model3, TestX);\n\n PreY = [label1, label2, label3];<\/code><\/pre>\n","protected":false},"excerpt":{"rendered":"
                \u672c\u6587\u4e3b\u8981\u4ecb\u7ecd\u5728Matlab\u4e2d\u4f7f\u7528\u5b98\u65b9\u5e93\u51fd\u6570\u63d0\u4f9b\u7684SVM\u652f\u6301\u5411\u91cf\u673a\u548cNNW\u795e\u7ecf\u7f51\u7edc\uff0c\u540c\u65f6\u5c06\u8fd9\u4e24\u79cd\u6a21\u578b\u5e94\u7528\u5230\u591a …<\/p>","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[65],"tags":[],"amp_enabled":true,"_links":{"self":[{"href":"https:\/\/swordofmorning.com\/index.php\/wp-json\/wp\/v2\/posts\/251"}],"collection":[{"href":"https:\/\/swordofmorning.com\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/swordofmorning.com\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/swordofmorning.com\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/swordofmorning.com\/index.php\/wp-json\/wp\/v2\/comments?post=251"}],"version-history":[{"count":1,"href":"https:\/\/swordofmorning.com\/index.php\/wp-json\/wp\/v2\/posts\/251\/revisions"}],"predecessor-version":[{"id":437,"href":"https:\/\/swordofmorning.com\/index.php\/wp-json\/wp\/v2\/posts\/251\/revisions\/437"}],"wp:attachment":[{"href":"https:\/\/swordofmorning.com\/index.php\/wp-json\/wp\/v2\/media?parent=251"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/swordofmorning.com\/index.php\/wp-json\/wp\/v2\/categories?post=251"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/swordofmorning.com\/index.php\/wp-json\/wp\/v2\/tags?post=251"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}