{"id":4619,"date":"2020-12-05T08:25:02","date_gmt":"2020-12-04T23:25:02","guid":{"rendered":"http:\/\/hituji1012.xsrv.jp\/?p=391"},"modified":"2026-08-01T13:56:23","modified_gmt":"2026-08-01T04:56:23","slug":"240","status":"publish","type":"post","link":"https:\/\/tetsumag.com\/blog\/2020\/12\/05\/240\/","title":{"rendered":"\u3010Python\u3011\u30ed\u30b8\u30b9\u30c6\u30a3\u30c3\u30af\u56de\u5e30\u306e\u308f\u304b\u308a\u3084\u3059\u3044\u89e3\u8aac"},"content":{"rendered":"\n

<\/strong><\/p>\n

\u7dda\u5f62\u56de\u5e30\u304c\u7406\u89e3\u3067\u304d\u305f\u3089\u3001\u6b21\u306f\u30ed\u30b8\u30b9\u30c6\u30a3\u30c3\u30af\u56de\u5e30\u306b\u3064\u3044\u3066\u898b\u3066\u3044\u304d\u307e\u3059\u3002<\/p>\n

\u30ed\u30b8\u30b9\u30c6\u30a3\u30c3\u30af\u56de\u5e30\u306f\u5206\u985e\u30e2\u30c7\u30eb\u306b\u4f7f\u7528\u3055\u308c\u3001\u826f\u3044\/\u60aa\u3044\u306a\u3069\u306e0\/1\u5224\u5b9a\u3092\u3059\u308b\u6a5f\u68b0\u5b66\u7fd2\u3067\u826f\u304f\u4f7f\u7528\u3055\u308c\u307e\u3059\u3002<\/p>\n

\u4f8b\u3048\u3070\u3001\u764c\u306e\u5224\u5b9a\u3084\u3001\u8ff7\u60d1\u30e1\u30fc\u30eb\u306e\u5224\u5b9a\u306a\u3069\u304c\u3088\u304f\u4f8b\u3067\u51fa\u3066\u304d\u307e\u3059\u3002<\/p>\n

\u7dda\u5f62\u56de\u5e30\u306b\u30b7\u30b0\u30e2\u30a4\u30c9\u95a2\u6570\u3068\u3044\u3046\u3082\u306e\u304c\u65b0\u3057\u304f\u5165\u308b\u306e\u3067\u3001\u5148\u306b\u7dda\u5f62\u56de\u5e30\u3092\u7406\u89e3\u3057\u3066\u304a\u304f\u3053\u3068\u3092\u304a\u52e7\u3081\u3057\u307e\u3059\u3002<\/p>\n

\u96e3\u3057\u3044\u6570\u5b66\u306e\u77e5\u8b58\u306f\u3067\u304d\u308b\u3060\u3051\u7701\u3044\u3066\u3067\u304d\u308b\u3060\u3051\u7c21\u5358\u306b\u8aac\u660e\u3057\u307e\u3059\u3002<\/p>\n

\u305f\u3060\u3001\u4e2d\u76e4\u4ee5\u964d\u306fpython\u306e\u57fa\u790e\u7684\u306a\u77e5\u8b58\u306f\u5fc5\u8981\u3067\u3059\u3002<\/p>\n

\n
\n

\u76ee\u6b21<\/p>\n

  • \u30ed\u30b8\u30b9\u30c6\u30a3\u30c3\u30af\u56de\u5e30<\/a>
    • \u30b7\u30b0\u30e2\u30a4\u30c9\u95a2\u6570\u306fS\u5b57\u578b<\/a><\/li>
    • \u7dda\u5f62\u56de\u5e30\u3068\u30b7\u30b0\u30e2\u30a4\u30c9\u95a2\u6570\u3092\u7d44\u307f\u5408\u308f\u305b\u30660\/1\u306e\u5224\u5b9a\u30e2\u30c7\u30eb\u3092\u4f5c\u308b<\/a><\/li>
    • \u30ed\u30b8\u30b9\u30c6\u30a3\u30c3\u30af\u56de\u5e30\u306eloss(\u640d\u5931\u95a2\u6570)<\/a><\/li>
    • loss(\u640d\u5931\u95a2\u6570)\u306e\u9078\u5b9a\u306f\u6a5f\u68b0\u5b66\u7fd2\u306e\u30e2\u30c7\u30eb\u306b\u5408\u308f\u305b\u308c\u3070\u3088\u3044<\/a><\/li><\/ul><\/li>
    • \u30ed\u30b8\u30b9\u30c6\u30a3\u30c3\u30af\u56de\u5e30\u30670\/1\u306e\u5206\u985e\u3092\u3059\u308b<\/a>
      • scikit learn\u306e\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb<\/a><\/li>
      • \u30e9\u30a4\u30d6\u30e9\u30ea\u306e\u8aad\u307f\u8fbc\u307f<\/a><\/li>
      • \u30c7\u30fc\u30bf\u306e\u6e96\u5099\u3068\u78ba\u8a8d<\/a><\/li>
      • \u5b66\u7fd2<\/a><\/li>
      • \u7d50\u679c\u78ba\u8a8d<\/a><\/li><\/ul><\/li>
      • \u307e\u3068\u3081<\/a><\/li><\/ul><\/nav><\/div>\n

        <\/span>\u30ed\u30b8\u30b9\u30c6\u30a3\u30c3\u30af\u56de\u5e30<\/span><\/h2>\n

        0\u304b1\u306b\u5206\u3051\u308b\u65b9\u6cd5\u3092\u4e8c\u9805\u5206\u985e<\/strong>\u3068\u8a00\u3044\u307e\u3059\u3002<\/p>\n

        \u6a5f\u68b0\u5b66\u7fd2\u30670\u304b1\u3092\u5224\u5b9a\u3055\u305b\u308b\u306b\u306f\u3069\u306e\u3088\u3046\u306a\u30e2\u30c7\u30eb\u304c\u5fc5\u8981\u3067\u3057\u3087\u3046\u304b\uff1f<\/p>\n

        \u3053\u3053\u3067\u306f\u307e\u305a\u30b7\u30b0\u30e2\u30a4\u30c9\u95a2\u6570\u306b\u3064\u3044\u3066\u8a73\u3057\u304f\u898b\u3066\u3044\u304d\u307e\u3057\u3087\u3046\u3002<\/p>\n

        <\/span>\u30b7\u30b0\u30e2\u30a4\u30c9\u95a2\u6570\u306fS\u5b57\u578b<\/span><\/h3>\n

        \u30ed\u30b8\u30b9\u30c6\u30a3\u30c3\u30af\u56de\u5e30\u3067\u306f\u3001\u4e0b\u8a18\u306e\u3088\u3046\u306bx\u3092\u4e0e\u3048\u308b\u3068\u30010\u301c1\u306e\u9593\u3067\u5909\u5316\u3059\u308b\u30b7\u30b0\u30e2\u30a4\u30c9\u95a2\u6570\u3068\u3044\u3046\u306e\u304c\u3042\u308b\u306e\u3067\u3001\u305d\u308c\u3092\u4f7f\u3063\u3066\u3044\u304d\u307e\u3059\u3002<\/p>\n

        \u30b7\u30b0\u30e2\u30a4\u30c9\u95a2\u6570\u304b\u3089\u3067\u304d\u305f\u5024\u304c0.5\u4ee5\u4e0a\u3067\u3042\u308c\u3070\u3001\u300c\uff11\u300d\u305d\u308c\u4ee5\u5916\u306a\u3089\u300c\uff10\u300d\u3068\u3044\u3046\u3053\u3068\u30670\/1\u5224\u5b9a\u304c\u3067\u304d\u307e\u3059\u3002<\/p>\n

        $$ f(x) = \\frac{1}{1+e^{(-x)}} $$<\/p>\n

        \"\"<\/a><\/p>\n

        \u516c\u5f0f\u306f\u899a\u3048\u306a\u304f\u3066\u3082\u3001\u3053\u3046\u3044\u3046S\u5b57\u578b\u306b\u306a\u308b\u5f0f\u3068\u3044\u3046\u304f\u3089\u3044\u3067\u826f\u3044\u3067\u3059\u3002<\/span><\/p>\n

        \u30b7\u30b0\u30e2\u30a4\u30c9\u95a2\u6570\u3067\u5f97\u3089\u308c\u308b\u5024\u3092\u78ba\u7387\u3068\u3057\u3066\u6349\u3048\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/p>\n

        \u4f8b\u3048\u3070\u3001\u30b7\u30b0\u30e2\u30a4\u30c9\u95a2\u6570\u304b\u30890.7\u3068\u3044\u3046\u7d50\u679c\u304c\u5f97\u3089\u308c\u305f\u3089\u300170%\u3067\u764c\u3067\u3042\u308b\u3068\u3044\u3063\u305f\u5224\u5b9a\u304c\u3059\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/p>\n

        \u3067\u3059\u306e\u3067\u30010.5\u4ee5\u4e0a\u304c1\u3067\u306f\u306a\u304f\u3001\u3088\u308a\u53b3\u3057\u304f0.8\u4ee5\u4e0a\u306a\u30891\u3068\u3044\u3046\u3053\u3068\u3082\u3067\u304d\u307e\u3059\u3002<\/p>\n

        python\u3067\u306e\u30b3\u30fc\u30c9\u306f\u4e0b\u8a18\u306e\u3088\u3046\u306a\u30b7\u30f3\u30d7\u30eb\u306a\u3082\u306e\u3067\u3059\u3002<\/p>\n

        import math\r\ndef sigmoid(x):\r\n    return 1 \/ (1 + math.e**-x)<\/pre>\n
        -10\u301c10\u3067\u30d7\u30ed\u30c3\u30c8\u3057\u3066\u307f\u307e\u3057\u3087\u3046\u3002<\/p>\n
        import matplotlib.pyplot as plt\r\nx = range(-10,10)\r\ny = [sigmoid(d) for d in x]\r\nplt.scatter(x, y)<\/pre>\n
        \"\"<\/a><\/p>\n
        \u671f\u5f85\u901a\u308a\u306e\u30d7\u30ed\u30c3\u30c8\u3067\u3059\u3002<\/p>\n
        x \u3068 y\u306e\u30c7\u30fc\u30bf\u3092\u4f5c\u308b\u3068\u3053\u308d\u306f\u4e0b\u8a18\u30b3\u30fc\u30c9\u3067\u3082\u540c\u3058\u3067\u3059\u306e\u3067\u3001\u5206\u304b\u308a\u3084\u3059\u3044\u65b9\u3092\u4f7f\u3063\u3066\u304f\u3060\u3055\u3044\u3002<\/p>\n
        y\u306b\u306f\u7a7a\u306e\u30ea\u30b9\u30c8\u3092\u7528\u610f\u3057\u3001append\u30e1\u30bd\u30c3\u30c9\u3067\u30ea\u30b9\u30c8\u3092\u8ffd\u52a0\u3057\u3066\u3044\u304d\u307e\u3059\u3002<\/p>\n
        x = range(-10,10)\r\ny = []\r\nfor d in x:\r\n    y.append(sigmoid(d))<\/pre>\n
        \u53c2\u8003\u306e\u305f\u3081\u306b\u3001\u30b7\u30b0\u30e2\u30a4\u30c9\u95a2\u6570\u306e\u5b9f\u88c5\u3092\u7d39\u4ecb\u3057\u307e\u3057\u305f\u304c\u3001\u81ea\u3089\u5b9f\u88c5\u3059\u308b\u3053\u3068\u306f\u307b\u307c\u306a\u3044\u306e\u3067\u3001\u7c21\u5358\u306a\u3093\u3060\u3068\u3044\u3046\u3053\u3068\u304c\u308f\u304b\u3063\u3066\u3044\u305f\u3060\u3051\u308c\u3070\u826f\u3044\u3067\u3059\u3002<\/p>\n
        \u3010Python\u3011list\u306e\u4f7f\u3044\u65b9<\/a><\/p><\/blockquote>\n