{"id":131,"date":"2014-07-03T00:25:10","date_gmt":"2014-07-02T15:25:10","guid":{"rendered":"https:\/\/www.dogrow.net\/nnet\/?p=131"},"modified":"2025-06-11T22:49:57","modified_gmt":"2025-06-11T13:49:57","slug":"blog10","status":"publish","type":"post","link":"https:\/\/www.dogrow.net\/nnet\/blog10\/","title":{"rendered":"(10) cuda-convnet\u306e\u30ec\u30dd\u30fc\u30c8\u8868\u793a"},"content":{"rendered":"

(9) cuda-convnet\u3092\u8a66\u3057\u3066\u307f\u308b<\/a> \u306e\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u7d50\u679c\u3092\u30ec\u30dd\u30fc\u30c8\u8868\u793a\u3055\u305b\u3066\u307f\u307e\u3059\u3002
\n\u3053\u308c\u3082 cuda-convnet<\/span> \u304c\u63d0\u4f9b\u3057\u3066\u3044\u308b\u6a5f\u80fd\u3067\u3059\u3002<\/p>\n

1. cost function\u306e\u51fa\u529b\u5024<\/h1>\n
[user@linux]$ SAVEDATA=..\/save\/ConvNet__2014-07-02_23.29.41\r\n[user@linux]$ python shownet.py -f $SAVEDATA --show-cost=logprob\r\n<\/pre>\n
\"20140702_05\"<\/a><\/p>\n
2. error rate<\/h1>\n
[user@linux]$ python shownet.py -f $SAVEDATA --show-cost=logprob --cost-idx=1\r\n<\/pre>\n
\"20140702_06\"<\/a><\/p>\n
3. learned filters (conv1 layer)<\/h1>\n
conv1<\/span> \u3068\u306f\u3001\u30cd\u30c3\u30c8\u30ef\u30fc\u30af\u5b9a\u7fa9\u30d5\u30a1\u30a4\u30eb\u306e\u4e2d\u3067\u5f53\u8a72\u30ec\u30a4\u30e4\u30fc\u306b\u4ed8\u3051\u305f\u540d\u524d\u3067\u3059\u3002
\n\u5148\u306e\u5b9f\u884c\u6642\u306b\u6307\u5b9a\u3057\u305f\u30cd\u30c3\u30c8\u30ef\u30fc\u30af\u5b9a\u7fa9\u30d5\u30a1\u30a4\u30eb\u306f\u3053\u308c\u3067\u3059\u3002
\n –layer-def=.\/example-layers\/layers-19pct.cfg<\/span><\/p>\n
[user@linux]$ python shownet.py -f $SAVEDATA --show-filters=conv1\r\n<\/pre>\n
\"20140702_07\"<\/a><\/p>\n
4. learned filters (conv2 layer)<\/h1>\n
[user@linux]$ python shownet.py -f $SAVEDATA --show-filters=conv2\r\n<\/pre>\n
\"20140702_08\"<\/a><\/p>\n
5. learned filters (conv3 layer)<\/h1>\n
[user@linux]$ python shownet.py -f $SAVEDATA --show-filters=conv3\r\n<\/pre>\n
\"20140702_09\"<\/a><\/p>\n
6. learned filters (fc10 layer)<\/h1>\n
–channels=64<\/span> \u306f\u3001\u76f4\u524d\u30ec\u30a4\u30e4\u30fc(pool3)\u306e channel\u6570\u3092\u6307\u5b9a\u3057\u307e\u3059\u3002<\/p>\n
[user@linux]$ python shownet.py -f $SAVEDATA --show-filters=fc10 --channels=64\r\n<\/pre>\n
\"20140702_10\"<\/a><\/p>\n
7. \u5206\u985e\u7d50\u679c\u306e\u3046\u3061\uff18\u500b\u3092\u30e9\u30f3\u30c0\u30e0\u8868\u793a<\/h1>\n
[user@linux]$ python shownet.py -f $SAVEDATA --show-preds=probs\r\n<\/pre>\n
\u6b63\u89e3\u3057\u305f\u3082\u306e\u306f\u8d64\u3044\u7dda<\/span>\u3001\u4e0d\u6b63\u89e3\u306e\u3082\u306e\u306f\u9752\u3044\u7dda <\/span>\u3067\u8868\u793a\u3055\u308c\u307e\u3059\u3002
\n\u7dda\u306e\u9577\u3055\u306f\u51fa\u529b\u30e6\u30cb\u30c3\u30c8\u306e\u51fa\u529b\u5024\u306e\u306f\u305a…
\n\"20140702_11\"<\/a><\/p>\n
\u30e9\u30f3\u30c0\u30e0\u306a\u306e\u3067\u6bce\u56de\u8868\u793a\u3055\u308c\u308b\u753b\u50cf\u304c\u5909\u308f\u308a\u307e\u3059\u3002
\n\"20140702_12\"<\/a><\/p>\n
8. \u5206\u985e\u7d50\u679c\u306e\u3046\u3061\uff18\u500b\u3092\u30e9\u30f3\u30c0\u30e0\u8868\u793a(\u30a8\u30e9\u30fc\u306e\u307f)<\/h1>\n
–show-preds<\/span> \u306b\u52a0\u3048\u3066 –only-errors=1<\/span> \u3092\u6307\u5b9a\u3059\u308b\u3068\u3001\u8aa4\u8a8d\u8b58\u3057\u305f\u7d50\u679c\u306e\u307f\u3092\u8868\u793a\u3067\u304d\u307e\u3059\u3002<\/p>\n
[user@linux]$ python shownet.py -f $SAVEDATA --show-preds=probs --only-errors=1\r\n<\/pre>\n
\"20140702_13\"<\/a><\/p>\n
\n","protected":false},"excerpt":{"rendered":"
(9) cuda-convnet\u3092\u8a66\u3057\u3066\u307f\u308b \u306e\u30c8\u30ec\u30fc\u30cb\u30f3\u30b0\u7d50\u679c\u3092\u30ec\u30dd\u30fc\u30c8\u8868\u793a\u3055\u305b\u3066\u307f\u307e\u3059\u3002 \u3053\u308c\u3082 cuda-convnet \u304c\u63d0\u4f9b\u3057\u3066\u3044\u308b\u6a5f\u80fd\u3067\u3059\u3002 1. cost function\u306e\u51fa\u529b\u5024 [user@linux]\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":[8,10,6],"tags":[],"class_list":["post-131","post","type-post","status-publish","format-standard","hentry","category-cuda","category-cuda-convnet","category-6"],"views":2662,"amp_enabled":true,"_links":{"self":[{"href":"https:\/\/www.dogrow.net\/nnet\/wp-json\/wp\/v2\/posts\/131","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=131"}],"version-history":[{"count":16,"href":"https:\/\/www.dogrow.net\/nnet\/wp-json\/wp\/v2\/posts\/131\/revisions"}],"predecessor-version":[{"id":2542,"href":"https:\/\/www.dogrow.net\/nnet\/wp-json\/wp\/v2\/posts\/131\/revisions\/2542"}],"wp:attachment":[{"href":"https:\/\/www.dogrow.net\/nnet\/wp-json\/wp\/v2\/media?parent=131"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.dogrow.net\/nnet\/wp-json\/wp\/v2\/categories?post=131"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.dogrow.net\/nnet\/wp-json\/wp\/v2\/tags?post=131"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}