{"id":2599,"date":"2025-06-16T00:37:12","date_gmt":"2025-06-15T15:37:12","guid":{"rendered":"https:\/\/www.dogrow.net\/nnet\/?p=2599"},"modified":"2025-06-16T10:08:22","modified_gmt":"2025-06-16T01:08:22","slug":"blog61-%e3%83%8d%e3%82%a4%e3%83%94%e3%82%a2%e6%95%b0e%e3%81%af%e5%be%ae%e5%88%86%e3%81%97%e3%81%a6%e3%82%82e%e3%81%ab%e3%81%aa%e3%82%8b%e3%80%82","status":"publish","type":"post","link":"https:\/\/www.dogrow.net\/nnet\/blog61-%e3%83%8d%e3%82%a4%e3%83%94%e3%82%a2%e6%95%b0e%e3%81%af%e5%be%ae%e5%88%86%e3%81%97%e3%81%a6%e3%82%82e%e3%81%ab%e3%81%aa%e3%82%8b%e3%80%82\/","title":{"rendered":"(61) \u6a5f\u68b0\u5b66\u7fd2\u3067\u591a\u7528\u3055\u308c\u308b\u30cd\u30a4\u30d4\u30a2\u6570\u3068\u306f\uff1f"},"content":{"rendered":"

\u30101\u3011\u3084\u308a\u305f\u3044\u3053\u3068<\/h1>\n

\u6a5f\u68b0\u5b66\u7fd2\u3067\u306f\u3001\u6d3b\u6027\u5316\u95a2\u6570\u3001\u640d\u5931\u95a2\u6570\u3001\u78ba\u7387\u5206\u5e03\u3001\u5b66\u7fd2\u7387\u6e1b\u8870\u3001\u6b63\u5247\u5316\u3001etc. \u3044\u308d\u3044\u308d\u306a\u5834\u9762\u3067 e<\/span> \u3092\u4f7f\u3046\u3002
\n\u6a5f\u68b0\u5b66\u7fd2\u30e9\u30a4\u30d6\u30e9\u30ea\u3092\u4f7f\u308f\u305a\u306b\u3001\u6a5f\u68b0\u5b66\u7fd2\u30d7\u30ed\u30b0\u30e9\u30e0\u3092\u81ea\u529b\u5b9f\u88c5\u3059\u308b\u305f\u3081\u306b\u306f\u7d76\u5bfe\u306b\u5fc5\u8981\u306a\u77e5\u8b58\u306a\u306e\u3067\u3001\u6539\u3081\u3066\u6574\u7406\u3057\u3066\u304a\u304d\u305f\u3044\u3002<\/span><\/p>\n

\u6a5f\u68b0\u5b66\u7fd2\u30e9\u30a4\u30d6\u30e9\u30ea\u3092\u4f7f\u308f\u306a\u3044 Python\u81ea\u529b\u5b9f\u88c5\u30b7\u30ea\u30fc\u30ba<\/span> \u306f\u4ee5\u4e0b\u306e\u901a\u308a\u3002
\n(56) \u9006\u4f1d\u64ad\u3067\u4f7f\u3046 tanh(x)\u306e\u5fae\u5206<\/a>
\n(57) \u9006\u4f1d\u64ad\u3067\u4f7f\u3046 sigmoid(x)\u306e\u5fae\u5206<\/a>
\n(58) \u9006\u4f1d\u64ad\u3067\u4f7f\u3046 ReLU(x)\u306e\u5fae\u5206<\/a>
\n(59) \u9006\u4f1d\u64ad\u3067\u4f7f\u3046 MSE\uff08\u5e73\u5747\u4e8c\u4e57\u8aa4\u5dee\uff09\u306e\u5fae\u5206<\/a>
\n(60) \u4ed6\u30af\u30e9\u30b9\u5206\u985e\u3067\u4f7f\u3046 Softmax<\/a>
\n(61) \u6a5f\u68b0\u5b66\u7fd2\u3067\u591a\u7528\u3055\u308c\u308b\u30cd\u30a4\u30d4\u30a2\u6570\u3068\u306f\uff1f<\/a> \u2190\u4eca\u56de<\/span><\/p>\n

\u30102\u3011\u30cd\u30a4\u30d4\u30a2\u6570\u306e\u5b9a\u7fa9<\/h1>\n

(1) \u540d\u79f0<\/h3>\n

\u30cd\u30a4\u30d4\u30a2\u6570\uff08Napier’s constant\uff09
\n\u30aa\u30a4\u30e9\u30fc\u6570\uff08Euler’s number\uff09<\/p>\n

(2) \u540d\u79f0\u306e\u7531\u6765<\/h3>\n

\u30cd\u30a4\u30d4\u30a2\uff08John Napier\uff09\u306f\u5bfe\u6570\u306e\u6982\u5ff5\u3092\u767a\u660e\u3057\u305f\u4eba\u7269\u306e\u540d\u524d\u3002<\/p>\n

\u305f\u3060\u3057\u3001\u73fe\u5728\u306e\u81ea\u7136\u5bfe\u6570\uff08\u5e95\u304c e \u306e\u5bfe\u6570\uff09\u3068\u3044\u3046\u5f62\u3067\u306e\u5b9a\u7fa9\u30fb\u5229\u7528\u306f\u3001\u30aa\u30a4\u30e9\u30fc\uff08Leonhard Euler\uff09 \u306b\u3088\u3063\u3066\u4f53\u7cfb\u5316\u3055\u308c\u305f\u3002
\n\u2192 \u30aa\u30a4\u30e9\u30fc\u6570\uff08Euler’s number\uff09\u3068\u3082\u547c\u3070\u308c\u308b\u3002<\/p>\n

(3) \u30cd\u30a4\u30d4\u30a2\u6570\u306e\u5b9a\u7fa9<\/h3>\n

\u6975\u9650\u3092\u4f7f\u3063\u305f\u5b9a\u7fa9\uff1a<\/span>
\n\"\"<\/p>\n

\u7121\u9650\u7d1a\u6570\u3092\u4f7f\u3063\u305f\u5b9a\u7fa9\uff1a<\/span>
\n\"\"<\/p>\n

\u5b9f\u969b\u306b\u8a08\u7b97\u3055\u308c\u305f\u5024\u306f\uff1f
\ne = 2.718281828459045235360287471352…<\/span><\/p>\n

\u3053\u306e\u5024\u306b\u306a\u308b\u3053\u3068\u3092\u3001\u53ce\u675f\u30b0\u30e9\u30d5\u3067\u8996\u899a\u5316\u3057\u3066\u307f\u308b\u3002
\n\u3069\u3061\u3089\u306e\u5b9a\u7fa9\u5f0f\u3067\u3082 n \u2192 \u221e \u306e\u3068\u304d\u306b e \u3078\u53ce\u675f\u3059\u308b\u69d8\u5b50\u304c\u3088\u304f\u308f\u304b\u308b\u3002<\/p>\n

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

\u30103\u3011\u6a5f\u68b0\u5b66\u7fd2\u3067\u306f\u30cd\u30a4\u30d4\u30a2\u6570 e \u3092\u591a\u7528\u3059\u308b\u3002<\/h1>\n

\u6a5f\u68b0\u5b66\u7fd2\u3067\u306f\u3001\u6d3b\u6027\u5316\u95a2\u6570\u3001\u640d\u5931\u95a2\u6570\u3001\u78ba\u7387\u5206\u5e03\u3001\u5b66\u7fd2\u7387\u6e1b\u8870\u3001\u6b63\u5247\u5316\u3001etc. \u3044\u308d\u3044\u308d\u306a\u5834\u9762\u3067 e\u3092\u4f7f\u3046\u3002<\/p>\n

\u305d\u308c\u306f\u30fb\u30fb\u30fb
\n\u6570\u5b66\u7684\u30fb\u8a08\u7b97\u7684\u306b\u3068\u3066\u3082\u90fd\u5408\u304c\u826f\u3044\u304b\u3089\u3002<\/span>
\n\u30fb\u6307\u6570\u95a2\u6570\u3068\u5fae\u5206\u306e\u76f8\u6027\u304c\u826f\u304f\uff08\uff1d\u5e95\u304ce\u3060\u3068\u5fae\u5206\u3057\u305f\u5024\u3082\u540c\u3058\uff09\u5fae\u5206\u304c\u30b7\u30f3\u30d7\u30eb\u306b\u306a\u308b\u3002\u52fe\u914d\u8a08\u7b97\u304c\u7c21\u5358\u3067\u5b89\u5b9a\u3059\u308b\u3002
\n\u30fb\u6307\u6570\u95a2\u6570\u3068\u5bfe\u6570\u95a2\u6570\u304c\u5b8c\u5168\u306b\u9006\u95a2\u6570\u3002Softmax\u95a2\u6570 \u3084 \u4ea4\u5dee\u30a8\u30f3\u30c8\u30ed\u30d4\u30fc\u640d\u5931\u95a2\u6570 \u306b\u304a\u3044\u3066\u5bfe\u6570\u30fb\u6307\u6570\u306e\u5909\u63db\u304c\u5bb9\u6613\u3002
\n\u30fbSoftmax + Cross Entropy\u306e\u7d44\u307f\u5408\u308f\u305b\u3067\u306f\u3001\u6307\u6570\u3068\u5bfe\u6570\u304c\u6253\u3061\u6d88\u3057\u5408\u3063\u3066\u8a08\u7b97\u304c\u697d\u306b\u306a\u308b\u3002
\n\u30fb\u30ed\u30b8\u30b9\u30c6\u30a3\u30c3\u30af\u56de\u5e30\u306e\u640d\u5931\u95a2\u6570\u304c\u51f8\uff08\u6700\u5c0f\u5024\u304c1\u3064\uff09\u306b\u306a\u308b\u3053\u3068\u304c\u591a\u3044\u3002
\n\u30fb\u78ba\u7387\u5206\u5e03\u306b\u304a\u3044\u3066 e\u306f\u9023\u7d9a\u7684\u306a\u78ba\u7387\u5909\u5316\u3092\u8868\u73fe\u3059\u308b\u306e\u306b\u5411\u3044\u3066\u3044\u308b\u3002
\n\u30fb\u5b66\u7fd2\u7387\u6e1b\u8870\uff08Exponential Decay\uff09\u3067\u4f7f\u3044\u3084\u3059\u3044\u3002
\n\u30fb\u6b63\u5247\u5316\uff08weight decay\uff09\u3082\u6307\u6570\u95a2\u6570\u306e\u6027\u8cea\u3092\u6d3b\u7528\u3002L2\u6b63\u5247\u5316\u306f e\u306e\u5fae\u5206\u7279\u6027\u304c\u3042\u308b\u304b\u3089\u7c21\u6f54\u306b\u8a08\u7b97\u3067\u304d\u308b\u3002<\/p>\n

\u5e95\u304c e \u3067\u306a\u304f\u3066\u3082\u6a5f\u68b0\u5b66\u7fd2\u3092\u5b9f\u73fe\u3059\u308b\u3053\u3068\u306f\u3067\u304d\u308b\u3002<\/span><\/p>\n

\u3057\u304b\u3057…
\n\u4f8b\u3048\u3070\u300110\u9032\u6570\u306b\u99b4\u67d3\u307f\u304c\u3042\u308b\u6211\u3005\u306b\u3068\u3063\u3066\u308f\u304b\u308a\u3084\u3059\u305d\u3046\u306a \u5e95=10 \u3068\u3057\u3066\u307f\u308b\u3002<\/p>\n

\u3059\u308b\u3068…
\n\"\"<\/p>\n

e \u3092\u4f7f\u3063\u305f\u65b9\u304c\u30b7\u30f3\u30d7\u30eb\u306b\u8a18\u8ff0\u3067\u304d\u308b\u3057\u3001\u8a08\u7b97\u30ed\u30b8\u30c3\u30af\u3082\u5358\u7d14\u306b\u306a\u308b\u3002
\nFPU\u304c\u7279\u6b8a\u306a\u56de\u8def\u3067\u3042\u3063\u305f\u6642\u4ee3\u306f\u9065\u304b\u6614\u3001\u8a08\u7b97\u901f\u5ea6\u3082\u554f\u984c\u306a\u3044\u3002<\/p>\n

\u30104\u3011e\u306e\u7279\u5fb4<\/h1>\n\n
#<\/th>\u7279\u5fb4<\/th><\/tr>
1<\/td>\u5fae\u5206\u3057\u3066\u3082\u5f62\u304c\u5909\u308f\u3089\u306a\u3044\u3002<\/span> \u6307\u6570\u95a2\u6570 ex<\/sup> \u306f\u5fae\u5206\u3057\u3066\u3082\u5909\u308f\u3089\u306a\u3044\uff08\u5b66\u7fd2\u306b\u4fbf\u5229\uff09<\/td><\/tr>
2<\/td>\u5bfe\u6570\u95a2\u6570\u3068\u6307\u6570\u95a2\u6570\u304c\u5b8c\u5168\u306b\u9006\u95a2\u6570\u306b\u306a\u308b\u3002<\/span> log(ex<\/sup>) = x \u306a\u306e\u3067\u3001\u8a08\u7b97\u304c\u30b7\u30f3\u30d7\u30eb<\/td><\/tr>
3<\/td>\u6210\u9577\u3084\u78ba\u7387\u306e\u5909\u5316\u3092\u306a\u3081\u3089\u304b\u306b\u8868\u73fe\u3067\u304d\u308b<\/td><\/tr>\n<\/table>\n

\u4e0a\u8a18\u306e\u5f0f\u3092 1\u679a\u306e\u30b0\u30e9\u30d5\u4e0a\u306b\u8868\u793a\u3057\u3066\u307f\u308b\u3002
\n\u305f\u3060\u3057\u3001dex<\/sup>\/dx = ex<\/sup> \u306a\u306e\u3067\u3001\u3053\u306e 2\u5f0f\u306f\u5b8c\u5168\u306b\u91cd\u306a\u3063\u3066\u3044\u308b\u3002<\/p>\n

\"\"<\/p>\n

\r\nimport numpy as np\r\nimport matplotlib.pyplot as plt\r\n\r\n# x\u306e\u7bc4\u56f2\u3092\u5b9a\u7fa9\r\nx = np.linspace(-2, 2, 400)\r\n\r\n# \u5404\u95a2\u6570\u306e\u5b9a\u7fa9\r\ny_exp = np.exp(x)             # y = e^x\r\ny_log_exp = np.log(np.exp(x)) # y = log(e^x) \u2192 x\r\ny_derivative = np.exp(x)      # y = d\/dx e^x \u2192 e^x\r\n\r\n# \u30b0\u30e9\u30d5\u306e\u63cf\u753b\r\nplt.figure(figsize=(8, 6))\r\nplt.plot(x, y_exp, label=r'$y = e^x$', color='blue')\r\nplt.plot(x, y_log_exp, label=r'$y = \\log(e^x)$', color='green')\r\nplt.plot(x, y_derivative, label=r'$y = \\frac{d}{dx}e^x$', color='red', linestyle='--')\r\n\r\nplt.title('Graphs of $y = e^x$, $y = \\\\log(e^x)$, and $y = \\\\frac{d}{dx}e^x$')\r\nplt.xlabel('x')\r\nplt.ylabel('y')\r\nplt.grid(True)\r\nplt.legend()\r\nplt.show()\r\n<\/pre>\n
\u30105\u3011\u300c\u5fae\u5206\u3057\u3066\u3082\u540c\u3058\u300d\u3092\u8a3c\u660e\u3059\u308b\u3002<\/h1>\n
1) ex<\/sup>\u306e\u5fae\u5206\u306f\uff1f<\/h2>\n
(1) \u76ee\u7684<\/h3>\n
\u4e0b\u5f0f\u3092\u8a3c\u660e\u3057\u305f\u3044\u3002
\n\"\"<\/p>\n
(2) \u8a3c\u660e<\/h3>\n
\u5fae\u5206\u306e\u5b9a\u7fa9\u306f\uff1f
\n\"\"<\/p>\n
\u3053\u308c\u306b f(x) = ex<\/sup> \u3092\u4ee3\u5165\u3059\u308b\u3002
\n\"\" \u30fb\u30fb\u30fb(1)<\/p>\n
\u6307\u6570\u6cd5\u5247\u306b\u3088\u308a\u3001
\n\"\"<\/p>\n
\u3053\u308c\u3092 (1)\u306b\u4ee3\u5165\u3059\u308b\u3002
\n\"\" \u30fb\u30fb\u30fb(2)<\/p>\n
\u3053\u3053\u3067\u3001
\n\"\" \u30fb\u30fb\u30fb(3)<\/p>\n
\u304c\u6210\u7acb\u3059\u308b\u306e\u3067\u3001\u3053\u308c\u3092 (2)\u306b\u4ee3\u5165\u3059\u308b\u3068\u3001
\n\"\"
\n\u8a3c\u660e\u3067\u304d\u305f\u3002<\/span><\/p>\n
(3) \u5fc5\u8981\u306a\u4e88\u5099\u77e5\u8b58<\/h3>\n
\u524d\u8ff0\u306e (3) \u306b\u3064\u3044\u3066\u8a3c\u660e\u3059\u308b\u3002
\n\"\"<\/p>\n
eh<\/sup>\u3092\u30c6\u30a4\u30e9\u30fc\u5c55\u958b\u3059\u308b\u3068\u3001
\n\"\"<\/p>\n
\u4e21\u8fba\u304b\u3089 1\u3092\u5f15\u304d\u3001h\u3067\u5272\u308b\u3068\u3001
\n\"\"<\/p>\n
\u3053\u3053\u3067 h\u3092 0\u306b\u8fd1\u3065\u3051\u308b\u3068\u3001
\n\"\"<\/p>\n
\u3088\u3063\u3066\u3001\u4e0b\u5f0f\u304c\u6210\u7acb\u3059\u308b\u3002
\n\"\"
\n\u8a3c\u660e\u3067\u304d\u305f\u3002<\/span><\/p>\n
\u30106\u3011\u53c2\u8003<\/h1>\n
1) 10x<\/sup>\u306e\u5fae\u5206<\/h2>\n
\u516c\u5f0f\u3092\u4f7f\u308f\u305a\u306b\u611a\u76f4\u306b\u5c55\u958b\u3057\u3066\u307f\u308b\u3002
\n\"\"<\/p>\n
\n","protected":false},"excerpt":{"rendered":"
\u30101\u3011\u3084\u308a\u305f\u3044\u3053\u3068 \u6a5f\u68b0\u5b66\u7fd2\u3067\u306f\u3001\u6d3b\u6027\u5316\u95a2\u6570\u3001\u640d\u5931\u95a2\u6570\u3001\u78ba\u7387\u5206\u5e03\u3001\u5b66\u7fd2\u7387\u6e1b\u8870\u3001\u6b63\u5247\u5316\u3001etc. \u3044\u308d\u3044\u308d\u306a\u5834\u9762\u3067 e \u3092\u4f7f\u3046\u3002 \u6a5f\u68b0\u5b66\u7fd2\u30e9\u30a4\u30d6\u30e9\u30ea\u3092\u4f7f\u308f\u305a\u306b\u3001\u6a5f\u68b0\u5b66\u7fd2\u30d7\u30ed\u30b0\u30e9\u30e0\u3092\u81ea\u529b\u5b9f\u88c5\u3059\u308b\u305f\u3081\u306b\u306f\u7d76\u5bfe\u306b\u5fc5\u8981\u306a\u77e5\u8b58\u306a\u306e\u3067\u2026 \u7d9a\u304d\u3092\u8aad\u3080 »<\/a><\/span><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[34,33,23],"tags":[],"class_list":["post-2599","post","type-post","status-publish","format-standard","hentry","category-34","category-33","category-23"],"views":809,"amp_enabled":true,"_links":{"self":[{"href":"https:\/\/www.dogrow.net\/nnet\/wp-json\/wp\/v2\/posts\/2599","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.dogrow.net\/nnet\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.dogrow.net\/nnet\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.dogrow.net\/nnet\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.dogrow.net\/nnet\/wp-json\/wp\/v2\/comments?post=2599"}],"version-history":[{"count":54,"href":"https:\/\/www.dogrow.net\/nnet\/wp-json\/wp\/v2\/posts\/2599\/revisions"}],"predecessor-version":[{"id":2933,"href":"https:\/\/www.dogrow.net\/nnet\/wp-json\/wp\/v2\/posts\/2599\/revisions\/2933"}],"wp:attachment":[{"href":"https:\/\/www.dogrow.net\/nnet\/wp-json\/wp\/v2\/media?parent=2599"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.dogrow.net\/nnet\/wp-json\/wp\/v2\/categories?post=2599"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.dogrow.net\/nnet\/wp-json\/wp\/v2\/tags?post=2599"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}