{"id":12138,"date":"2020-02-21T00:24:43","date_gmt":"2020-02-20T15:24:43","guid":{"rendered":"https:\/\/deeplearning.jp\/?page_id=12138"},"modified":"2020-03-18T15:33:58","modified_gmt":"2020-03-18T06:33:58","slug":"deep-generative-model-2019","status":"publish","type":"page","link":"https:\/\/deeplearning.jp\/en\/deep-generative-model-2019\/","title":{"rendered":"Deep Learning for NLP\u8b1b\u5ea7"},"content":{"rendered":"

\n
\n

Deep Learning for NLP \u30b5\u30de\u30fc\u30b9\u30af\u30fc\u30eb<\/h1>\n<\/div>\n<\/div>\n<\/div><\/div>

\u6982\u8981<\/h2>\n<\/div><\/div>
<\/div><\/div><\/div>
\u672c\u8b1b\u5ea7\u306f\u3001\u5b66\u751f\u3092\u5bfe\u8c61\u3068\u3057\u305f\u3001Deep Learning\u3092\u7528\u3044\u305f\u81ea\u7136\u8a00\u8a9e\u51e6\u7406\u306b\u7279\u5316\u3057\u305f\u51686\u56de\u30fb\u7d042\u9031\u9593\u306e\u30b5\u30de\u30fc\u30b9\u30af\u30fc\u30eb \u3067\u3059\u3002\u6771\u4eac\u5927\u5b66Deep Learning\u57fa\u790e\u8b1b\u5ea7\u30fb\u5fdc\u7528\u8b1b\u5ea7\u3092\u516c\u958b\u3057\u3066\u304d\u305f\u677e\u5c3e\u7814\u7a76\u5ba4\u304c\u5168\u9762\u7684\u306b\u6f14\u7fd2\u30b3\u30f3\u30c6\u30f3\u30c4\u3092\u76e3\u4fee\u30fb\u4f5c\u6210\u3057\u3066\u3044\u307e\u3059\u3002\u5b9f\u8df5\u7684\u306a\u6f14\u7fd2\u3092\u901a\u3057\u3066\u3001\u52b9\u7387\u7684\u306b\u5b66\u7fd2\u3092\u9032\u3081\uff0c\u624b\u3092\u52d5\u304b\u3057\u306a\u304c\u3089\u6280\u8853\u3092\u6df1\u304f\u7406\u89e3\u3057\uff0c\u5e45\u5e83\u3044\u30c8\u30d4\u30c3\u30af\u3092\u7db2\u7f85\u3057\u307e\u3059\u3002\u305d\u308c\u306b\u3088\u3063\u3066\u3001Deep Learning\u3092\u7528\u3044\u305f\u81ea\u7136\u8a00\u8a9e\u51e6\u7406\u306e\u7814\u7a76\u306e\u30b9\u30bf\u30fc\u30c8\u30e9\u30a4\u30f3\u306b\u7acb\u3066\u308b\u30ec\u30d9\u30eb\u306e\u77e5\u8b58\u30fb\u5b9f\u88c5\u529b\u306e\u7fd2\u5f97\u3092\u76ee\u6307\u3057\u307e\u3059\u3002\u5206\u6563\u8868\u73fe\u30fb\u8a00\u8a9e\u30e2\u30c7\u30eb\u304b\u3089\u59cb\u307e\u308a\u3001RNN\u30fbCNN\u30fbAttention\u3001\u751f\u6210\u30e2\u30c7\u30eb\u3001\u5f37\u5316\u5b66\u7fd2\u306b\u89e6\u308c\u3001\u73fe\u5728\u306e\u6982\u89b3\u3092\u5b66\u3076\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/div>\n<\/div><\/div><\/div><\/div><\/div>
<\/div><\/div><\/div><\/div>
<\/span><\/div><\/div><\/div><\/div><\/div>
<\/div><\/div><\/div>
\"\"<\/div>\n

1\u53d7\u8b1b\u8005 \u2013 1GPU\u30b5\u30fc\u30d0<\/h4>\n

\u5168\u3066\u306e\u53d7\u8b1b\u8005\u306b\u4e00\u4eba\u4e00\u53f0\u306e\u72ec\u7acb\u3057\u305f\u4eee\u60f3GPU\u30b5\u30fc\u30d0\u304c\u5272\u308a\u5f53\u3066\u3089\u308c\u307e\u3059\uff0e\u53d7\u8b1b\u4e2d\u306f\u4ed6\u306e\u53d7\u8b1b\u8005\u306e\u5f71\u97ff\u3092\u53d7\u3051\u308b\u3053\u3068\u306a\u304f\u5b66\u7fd2\u3092\u9032\u3081\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/p>\n<\/div><\/div><\/div><\/div><\/div>

\"\"<\/div>\n

\u30d6\u30e9\u30a6\u30b6\u306e\u307f\u5fc5\u8981<\/h4>\n

\u5fc5\u8981\u306a\u3082\u306e\u306fWeb\u30d6\u30e9\u30a6\u30b6\u306e\u307f\u3067\u3059\u3002\u30d6\u30e9\u30a6\u30b6\u3092\u901a\u3058\u3066GPU\u3092\u5229\u7528\u3057\u305fPython\u30d7\u30ed\u30b0\u30e9\u30df\u30f3\u30b0\u304c\u53ef\u80fd\u306a\u958b\u767a\u74b0\u5883\u300ciLect.net\u300d\u304c\u63d0\u4f9b\u3055\u308c\u307e\u3059\u3002\u30af\u30ea\u30c3\u30af\u64cd\u4f5c\u306e\u307f\u3067\u6f14\u7fd2\u3092\u958b\u59cb\u3059\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/p>\n<\/div><\/div><\/div><\/div><\/div>

\"\"<\/div>\n

\u69cb\u7bc9\u6e08\u307f\u74b0\u5883<\/h4>\n

\u5fc5\u8981\u306a\u5168\u3066\u306eDeep Learning\u30fb\u6a5f\u68b0\u5b66\u7fd2\u30e9\u30a4\u30d6\u30e9\u30ea\u304c\u30a4\u30f3\u30b9\u30c8\u30fc\u30eb\u3055\u308c\u3001\u69cb\u7bc9\u6e08\u307f\u306e\u74b0\u5883\u304c\u914d\u5e03\u3055\u308c\u307e\u3059\u306e\u3067\u3001\u3059\u3050\u306b\u5b66\u7fd2\u3092\u958b\u59cb\u3057\u672c\u984c\u306b\u96c6\u4e2d\u3057\u3066\u53d6\u308a\u7d44\u3080\u3053\u3068\u304c\u53ef\u80fd\u3067\u3059\u3002<\/p>\n<\/div><\/div><\/div><\/div><\/div><\/div>

\"\"<\/div>\n

\u6f14\u7fd2\u4e2d\u5fc3\u3067\u52b9\u7387\u826f\u304f\u77ed\u671f\u7fd2\u5f97<\/h4>\n

\u672c\u30d7\u30ed\u30b0\u30e9\u30e0\u306e\u6700\u5927\u306e\u7279\u5fb4\u306e\u4e00\u3064\u306f\u3001\u5168\u3066\u306e\u30c8\u30d4\u30c3\u30af\u306b\u3064\u3044\u3066\uff0c\u6f14\u7fd2\u3092\u4e2d\u5fc3\u306b\u69cb\u6210\u3055\u308c\u3066\u3044\u308b\u70b9\u3067\u3059\u3002\u5b9f\u969b\u306b\u624b\u3092\u52d5\u304b\u3057\u306a\u304c\u3089\u7406\u89e3\u3092\u9032\u3081\u308b\u3053\u3068\u3067\u3001\u52b9\u7387\u3088\u304f\u5b66\u7fd2\u3059\u308b\u3053\u3068\u304c\u3067\u304d\u307e\u3059\u3002<\/p>\n<\/div><\/div><\/div><\/div><\/div>

\"\"<\/div>\n

\u30e2\u30c7\u30eb\u306e\u5b66\u7fd2<\/h4>\n

Deep Learning\u6280\u8853\u3067\u306f\u3001\u30e2\u30c7\u30eb\u304c\u5b66\u7fd2\u3059\u308b\u69d8\u5b50\u3092\u89b3\u6e2c\u3057\u306a\u304c\u3089\u30d1\u30e9\u30e1\u30fc\u30bf\u3092\u8abf\u6574\u3059\u308b\u3053\u3068\u3067\u3001\u30a2\u30d7\u30ea\u30b1\u30fc\u30b7\u30e7\u30f3\u306b\u5fdc\u3058\u305f\u30d1\u30d5\u30a9\u30fc\u30de\u30f3\u30b9\u306e\u6700\u5927\u5316\u3092\u56f3\u308b\u3053\u3068\u304c\u975e\u5e38\u306b\u91cd\u8981\u3067\u3059\u3002\u3053\u306e\u4e00\u9023\u306e\u6d41\u308c\u3092\u5168\u3066\u306e\u6f14\u7fd2\u3067\u7d4c\u9a13\u3059\u308b\u3053\u3068\u3067\u3001\u91cd\u8981\u306a\u8981\u7d20\u3092\u8eab\u306b\u3064\u3051\u3089\u308c\u307e\u3059\u3002<\/p>\n<\/div><\/div><\/div><\/div><\/div>

\"\"<\/div>\n

\u677e\u5c3e\u7814\u7a76\u5ba4\u304c\u30b3\u30f3\u30c6\u30f3\u30c4\u4f5c\u6210<\/h4>\n

\u5b9f\u8df5\u578b\u306e\u30c7\u30fc\u30bf\u30b5\u30a4\u30a8\u30f3\u30c6\u30a3\u30b9\u30c8\u80b2\u6210\u8b1b\u5ea7\u304a\u3088\u3073Deep Learning\u8b1b\u5ea7\u30924\u5e74\u4ee5\u4e0a\u516c\u958b\u904b\u55b6\u3057\u3001\u306e\u3079\u5343\u4eba\u4ee5\u4e0a\u306e\u4eba\u6750\u3092\u80b2\u6210\u3057\u3066\u304d\u305f\u6771\u4eac\u5927\u5b66\u677e\u5c3e\u7814\u7a76\u5ba4\u304c\u30b3\u30f3\u30c6\u30f3\u30c4\u3092\u76e3\u4fee\u30fb\u958b\u767a\u3057\u307e\u3057\u305f\u3002<\/p>\n<\/div><\/div><\/div><\/div><\/div><\/div><\/div><\/div><\/div>

<\/div><\/div><\/div><\/div>
<\/span><\/div>

\u30b3\u30f3\u30c6\u30f3\u30c4<\/h2>\n<\/div><\/div><\/div><\/div><\/div><\/div>
<\/div><\/div><\/div>
<\/div>\n

\u7b2c1\u56de<\/span>NLP\u6982\u89b3<\/h4>\n

NLP\u3068\u306f, NLP\u306b\u304a\u3051\u308bDL\u306e\u5f79\u5272, \u30b1\u30fc\u30b9\u30b9\u30bf\u30c7\u30a3, Pytorch<\/p>\n<\/div><\/div><\/div><\/div><\/div>

<\/div>\n

\u7b2c2\u56de<\/span>\u5206\u6563\u8868\u73fe<\/h4>\n

\u5206\u6563\u8868\u73fe, \u65e5\u672c\u8a9e\u306e\u51e6\u7406, Softmax\u306e\u9ad8\u901f\u5316<\/p>\n<\/div><\/div><\/div><\/div><\/div>

<\/div>\n

\u7b2c3\u56de<\/span>\u8a00\u8a9e\u30e2\u30c7\u30eb<\/h4>\n

\u8a00\u8a9e\u30e2\u30c7\u30eb, RNN\u8a00\u8a9e\u30e2\u30c7\u30eb, Seq2Seq, \u53ef\u5909\u9577\u7cfb\u5217\u306e\u51e6\u7406, BeamSearch<\/p>\n<\/div><\/div><\/div><\/div><\/div><\/div>

<\/div>\n

\u7b2c4\u56de<\/span>\u5fdc\u7528\u2460\uff1aCNN\u30fbAttention<\/h4>\n

CNN, Attention,Transformer<\/p>\n<\/div><\/div><\/div><\/div><\/div>

<\/div>\n

\u7b2c5\u56de<\/span>\u5fdc\u7528\u2461\uff1a\u30c6\u30ad\u30b9\u30c8\u751f\u6210<\/h4>\n

VAE, GAN, Controllable Generation<\/p>\n<\/div><\/div><\/div><\/div><\/div>

<\/div>\n

\u7b2c6\u56de<\/span>\u5fdc\u7528\u2462\uff1a\u5f37\u5316\u5b66\u7fd2<\/h4>\n

Q\u5b66\u7fd2, Policy Gradient, SeqGAN, \u307e\u3068\u3081<\/p>\n<\/div><\/div><\/div><\/div><\/div><\/div><\/div><\/div><\/div>

<\/div><\/div><\/div><\/div>
<\/span><\/div>

\u4f01\u753b\u30fb\u904b\u55b6<\/h2>\n<\/div><\/div>