[B! chainer][ã‚ã¨ã§èªã‚€] deejayrokaã®ãƒ

deejayroka id:deejayroka

chainerã¨ã‚ã¨ã§èªã‚€ã«é–¢ã™ã‚‹deejayrokaã®ãƒ–ãƒƒã‚¯ãƒžãƒ¼ã‚¯ (4)

${{author_name}}$

{{{comment_expanded}}}

{{label}}

{{#is_bookmark}}ãƒªã‚¹ãƒˆ{{/is_bookmark}}{{^is_bookmark}}ãƒªãƒ³ã‚¯{{/is_bookmark}}

${{author_name}}$
{{author_name}}{{created}}
{{ #comment }}{{ comment }}{{ /comment }}
- {{ label }}

${{author_name}}$

{{{comment_expanded}}}

{{label}}

{{#is_bookmark}}ãƒªã‚¹ãƒˆ{{/is_bookmark}}{{^is_bookmark}}ãƒªãƒ³ã‚¯{{/is_bookmark}}

ã€ãƒ‡ã‚£ãƒ¼ãƒ—ãƒ©ãƒ¼ãƒ‹ãƒ³ã‚°ãƒ•ãƒ¬ãƒ¼ãƒ ãƒ¯ãƒ¼ã‚¯ã€‘Chainer, Keras, TensorFlow, Pytorch ã€æ›´æ–°2018/11/11ã€‘ - HELLO CYBERNETICS
ã¯ã˜ã‚ã« Chainer åˆ©ç‚¹æ¬ ç‚¹ æ„Ÿæƒ³ èª°ã«ã‚ªã‚¹ã‚¹ãƒ¡ã‹ TensorFlow åˆ©ç‚¹æ¬ ç‚¹ æ„Ÿæƒ³ èª°ã«ã‚ªã‚¹ã‚¹ãƒ¡ã‹ Keras åˆ©ç‚¹æ¬ ç‚¹ æ„Ÿæƒ³ èª°ã«ã‚ªã‚¹ã‚¹ãƒ¡ã‹ Pytorch ãªãœPytorchã‚’é¸ã‚“ã ã‹ PytorchãŒfast.aiã¨ãã®å—è¬›ç”Ÿã«ã‚‚ãŸã‚‰ã™åˆ©ç‚¹ èª°ã«ã‚ªã‚¹ã‚¹ãƒ¡ã‹ï¼ˆä½¿ã£ãŸã“ã¨ç„¡ã„ã®ã§åè¦‹ï¼‰ ãƒ•ãƒ¬ãƒ¼ãƒ ãƒ¯ãƒ¼ã‚¯æ¤œç´¢ãƒœãƒªãƒ¥ãƒ¼ãƒ ï¼ˆä¸€æ–°ã—ã¾ã—ãŸï¼‰ Define by Runã§æœ‰åãªChainerã¨PyTorchæ¯”è¼ƒ æ¯”è¼ƒç¯„å›²ï¼šæ—¥æœ¬ æ¯”è¼ƒç¯„å›²ï¼šå…¨ä¸–ç•Œ Chainerã€PyTorchã€TensorFlowã€Kerasã®æ¯”è¼ƒ æ¯”è¼ƒç¯„å›²ï¼šæ—¥æœ¬ æ¯”è¼ƒç¯„å›²ï¼šä¸–ç•Œ ã¯ã˜ã‚ã« ã“ã®è¨˜äº‹ã‚’å…¬é–‹ã—ã¦ã‹ã‚‰1å¹´ä»¥ä¸Šã®æœˆæ—¥ãŒçµŒã£ã¦ã„ã¾ã—ãŸã€‚ Googleæ¤œç´¢ã§ã€ã€ŒTensorFlowã€€PyTorchã€ã¨ã„ã†æ¤œç´¢ãƒ¯ãƒ¼ãƒ‰ã‚’å…¥ã‚Œã‚‹ã¨ãã‚Œãªã‚Šã®ä¸Šä½ã«ã“ã®è¨˜äº‹ãŒè¡¨ç¤ºã•ã‚Œã¦ã„ãŸã®ã§ã€ã‚ã¾ã‚Šã«æƒ…å ±ãŒå¤ã„ã¨ã„ã†ã“ã¨ã§æ›´æ–°ã™ã‚‹ã“ã¨ã«ã„
deejayroka 2017/09/25
chainer

ãƒ‡ã‚£ãƒ¼ãƒ—ãƒ©ãƒ¼ãƒ‹ãƒ³ã‚°

TensorFlow

ã‚ã¨ã§èªã‚€
ãƒªãƒ³ã‚¯
ãƒ‡ã‚£ãƒ¼ãƒ—ãƒ©ãƒ¼ãƒ‹ãƒ³ã‚°ã«ãŠã‘ã‚‹å¦ç¿’ã®ã‚³ãƒ„ã¨ãƒã‚¤ãƒ‘ãƒ¼ãƒ‘ãƒ©ãƒ¡ãƒ¼ã‚¿ã®ã‚¬ã‚¤ãƒ‰ãƒ©ã‚¤ãƒ³
ã¯ã˜ã‚ã« ä»Šæµè¡Œã‚Šã®ãƒ‡ã‚£ãƒ¼ãƒ—ãƒ©ãƒ¼ãƒ‹ãƒ³ã‚°(æ·±å±¤å¦ç¿’)ã§ã™ãŒã€ã„ã–è‡ªåˆ†ã§ã‚„ã£ã¦ã¿ã‚‹ã¨æ€ã£ãŸã‚ˆã†ã«ç²¾åº¦ãŒå‡ºãªã‹ã£ãŸã‚Šã€é‡ã¿ã®åˆæœŸåŒ–ã‚„å¦ç¿’çŽ‡ã‚’ã©ã®ã‚ˆã†ã«é¸æŠžã—ãŸã‚‰ã‚ˆã„ã®ã‹ã€ã¨ã„ã£ãŸã“ã¨ã§æˆ¸æƒ‘ã†ã“ã¨ãŒã‚ˆãã‚ã‚Šã¾ã™ã€‚ é‡ã¿ã®åˆæœŸåŒ–ã‚„å¦ç¿’çŽ‡ã¨ã„ã£ãŸã€ãƒ‹ãƒ¥ãƒ¼ãƒ©ãƒ«ãƒãƒƒãƒˆã§ã¯ãªãäººé–“ãŒé¸æŠžã—ãªã‘ã‚Œã°ã„ã‘ãªã„ãƒ‘ãƒ©ãƒ¡ãƒ¼ã‚¿ã‚’ãƒã‚¤ãƒ‘ãƒ¼ãƒ‘ãƒ©ãƒ¡ãƒ¼ã‚¿(Hyperparameter)ã¨å‘¼ã³ã¾ã™ã€‚é€†ã«è¨€ãˆã°ã€ãƒã‚¤ãƒ‘ãƒ¼ãƒ‘ãƒ©ãƒ¡ãƒ¼ã‚¿ã¨ã¯å¦ç¿’ã‚¢ãƒ«ã‚´ãƒªã‚ºãƒ ã«ã‚ˆã£ã¦æ±ºå®šã§ããšã€ãƒ’ãƒ¥ãƒ¼ãƒªã‚¹ãƒ†ã‚£ãƒƒã‚¯ãªéƒ¨åˆ†ã§ã‚‚ã‚ã‚Šã¾ã™ã€‚ ãƒã‚¤ãƒ‘ãƒ¼ãƒ‘ãƒ©ãƒ¡ãƒ¼ã‚¿ã¯è©¦è¡ŒéŒ¯èª¤ã§æ±ºå®šã—ã¦ã„ãã®ãŒåŸºæœ¬ã§ã™ãŒ(ã‚‚ã¡ã‚ã‚“è‡ªå‹•åŒ–æŠ€è¡“ã‚‚ã‚ã‚Šã¾ã™)ã€50å¹´ä»¥ä¸Šã®æ´å²ãŒã‚ã‚‹ãƒ‹ãƒ¥ãƒ¼ãƒ©ãƒ«ãƒãƒƒãƒˆã«ã¯çŸ¥æµã‚„ãƒ†ã‚¯ãƒ‹ãƒƒã‚¯ãŒå¤šãã®å…ˆäººã«ã‚ˆã£ã¦åŸ¹ã‚ã‚Œã¦ãã¾ã—ãŸã€‚ãã—ã¦ã€ä»Šã‚‚ãªãŠç ”ç©¶ãŒç››ã‚“ã«è¡Œã‚ã‚Œã¦ã„ã‚‹åˆ†é‡Žã§ã‚‚ã‚ã‚Šã¾ã™ã€‚ ã‚ã¨ã§å‚ç…§ã—ã‚„ã™ã„ã‚ˆã†ã«æœ¬è¨˜äº‹ã§ã¯ã€ãƒã‚¤ãƒ‘ãƒ¼ãƒ‘ãƒ©ãƒ¡ãƒ¼ã‚¿ã®æŒ‡å—ã‚’ã¾ã¨ã‚ã¦ã¿ã¾ã—ãŸã€‚
deejayroka 2017/08/22
ã‚ã¨ã§èªã‚€

å¦ç¿’

machinelearning

chainer

TensorFlow
ãƒªãƒ³ã‚¯
ã€PyTorchã€Chainerã€Kerasã€TensorFlowã€‘ãƒ‡ã‚£ãƒ¼ãƒ—ãƒ©ãƒ¼ãƒ‹ãƒ³ã‚°ã®ãƒ•ãƒ¬ãƒ¼ãƒ ãƒ¯ãƒ¼ã‚¯ã®åˆ©ç‚¹ãƒ»æ¬ ç‚¹ã€2017å¹´10æœˆæ›´æ–°ã€‘ - HELLO CYBERNETICS
ãƒ‡ã‚£ãƒ¼ãƒ—ãƒ©ãƒ¼ãƒ‹ãƒ³ã‚°ã®å¤§æµè¡Œã®ä¸ã€æ§˜ã€…ãªãƒ•ãƒ¬ãƒ¼ãƒ ãƒ¯ãƒ¼ã‚¯ãŒç™»å ´ã—ã€æ°—è»½ã«ãƒ—ãƒã‚°ãƒ©ãƒŸãƒ³ã‚°ãŒã§ãã‚‹ã‚ˆã†ã«ãªã‚Šã¾ã—ãŸã€‚ã—ã‹ã—ã€ãã‚“ãªä¸ã©ã®ãƒ•ãƒ¬ãƒ¼ãƒ ãƒ¯ãƒ¼ã‚¯ã‚’é¸ã¹ã°è‰¯ã„ã‹ã‚ã‹ã‚‰ãªã„ã¨ã„ã†äººã‚‚å¤šã„ã¨æ€ã„ã¾ã™ã€‚ãã‚“ãªäººã«å°‘ã—ã§ã‚‚å‚è€ƒã«ãªã‚Œã°ã¨æ€ã„è¨˜äº‹ã‚’æ›¸ãã¾ã™ã€‚ ã¯ã˜ã‚ã« Chainer ç‰¹å¾´ æŸ”è»Ÿãªè¨ˆç®—ã‚°ãƒ©ãƒ•ã®æ§‹ç¯‰ãŒå¯èƒ½ Pythonã«ã‚ˆã‚‹å®Ÿè£… ç›´æ„Ÿçš„ãªè¨ˆç®—ã‚°ãƒ©ãƒ•ã®æ§‹ç¯‰ãŒå¯èƒ½ ãƒ¡ãƒªãƒƒãƒˆãƒ»ãƒ‡ãƒ¡ãƒªãƒƒãƒˆ ãƒ¡ãƒªãƒƒãƒˆ ãƒ‡ãƒ¡ãƒªãƒƒãƒˆ ã¾ã¨ã‚ Keras ç‰¹å¾´ ã¨ã‚“ã§ã‚‚ãªãç°¡å˜ã«è¨ˆç®—ã‚°ãƒ©ãƒ•ã‚’è¨˜è¿°å¯èƒ½ é«˜é€Ÿè¨ˆç®—ãƒ©ã‚¤ãƒ–ãƒ©ãƒªã®ãƒ‡ã‚£ãƒ¼ãƒ—ãƒ©ãƒ¼ãƒ‹ãƒ³ã‚°ç”¨ãƒ©ãƒƒãƒ‘ãƒ¼ ã‚‚ã¯ã‚„ãƒ—ãƒã‚°ãƒ©ãƒŸãƒ³ã‚°ã®çµŒé¨“ã™ã‚‰ä¸è¦ ãƒ¡ãƒªãƒƒãƒˆãƒ»ãƒ‡ãƒ¡ãƒªãƒƒãƒˆ ãƒ¡ãƒªãƒƒãƒˆ ãƒ‡ãƒ¡ãƒªãƒƒãƒˆ ã¾ã¨ã‚ TensorFlow ç‰¹å¾´ åœ§å€’çš„ãªåˆ©ç”¨è€…æ•° ãƒ†ãƒ³ã‚½ãƒ«è¨ˆç®—ã‚’è¡Œã†ãƒ©ã‚¤ãƒ–ãƒ©ãƒª Define and Run è¿½åŠ ã®ãƒ©ã‚¤ãƒ–ãƒ©ãƒªãŒè±Šå¯Œ ãƒ¡ãƒªãƒƒãƒˆãƒ»ãƒ‡ãƒ¡ãƒªãƒƒãƒˆ ãƒ¡ãƒªãƒƒãƒˆ ãƒ‡ãƒ¡ãƒªãƒƒãƒˆ ã¾ã¨ã‚ PyT
deejayroka 2017/08/19
ã‚ã¨ã§èªã‚€

ã‚°ãƒ©ãƒ•

chainer

python

ãƒ‡ã‚£ãƒ¼ãƒ—ãƒ©ãƒ¼ãƒ‹ãƒ³ã‚°

ãƒ•ãƒ¬ãƒ¼ãƒ ãƒ¯ãƒ¼ã‚¯
ãƒªãƒ³ã‚¯
æ·±å±¤å¦ç¿’ã«ã‚ˆã‚‹è‡ªç„¶è¨€èªžå‡¦ç† - RNN, LSTM, ãƒ‹ãƒ¥ãƒ¼ãƒ©ãƒ«æ©Ÿæ¢°ç¿»è¨³ã®ç†è«–
æœ¬ç¨¿ã§ã¯ãƒ‹ãƒ¥ãƒ¼ãƒ©ãƒ«ãƒãƒƒãƒˆãƒ¯ãƒ¼ã‚¯ï¼Œèª¤å·®é€†ä¼æ’æ³•ï¼Œè¨€èªžãƒ¢ãƒ‡ãƒ«ï¼ŒRNNï¼ŒLSTMï¼Œãƒ‹ãƒ¥ãƒ¼ãƒ©ãƒ«æ©Ÿæ¢°ç¿»è¨³ã®ä¸€é€£ã®æ‰‹æ³•ã«ã¤ã„ã¦æ•°ç†çš„ã«è§£èª¬ã™ã‚‹ï¼Ž å‰ç·¨ã®ç›®æ¬¡ ãƒ‹ãƒ¥ãƒ¼ãƒ©ãƒ«ãƒãƒƒãƒˆãƒ¯ãƒ¼ã‚¯ é †ä¼æ’ (Forwardpropagation) é€†ä¼æ’ (Backpropagation) ãƒªã‚«ãƒ¬ãƒ³ãƒˆãƒ‹ãƒ¥ãƒ¼ãƒ©ãƒ«ãƒãƒƒãƒˆãƒ¯ãƒ¼ã‚¯ (RNN) Recurrent Neural Network Language Model (RNNLM) Backpropagation Through Time (BPTT) Long Short-Term Memory (LSTM) Gated Recurrent Unit (GRU) RNN ã®ãƒ‰ãƒãƒƒãƒ—ã‚¢ã‚¦ãƒˆã¨ãƒãƒƒãƒæ£è¦åŒ– ãƒ‹ãƒ¥ãƒ¼ãƒ©ãƒ«æ©Ÿæ¢°ç¿»è¨³ (NMT) Sequence to Sequence (seq2seq) æ³¨æ„ (Attention) åŒæ–¹å‘ã‚¨ãƒ³ã‚³ãƒ¼ãƒ€ãƒ¼ãƒ»å¤šå±¤LSTM è©•ä¾¡æ‰‹æ³•
deejayroka 2017/07/26
ã‚ã¨ã§èªã‚€

rnn

ãƒãƒƒãƒˆãƒ¯ãƒ¼ã‚¯

chainer

ãƒ¢ãƒ‡ãƒ«

Network

Deep Learning

AI

æ©Ÿæ¢°å¦ç¿’

deeplearning
ãƒªãƒ³ã‚¯
1