{"id":204,"date":"2018-11-24T12:53:15","date_gmt":"2018-11-24T05:53:15","guid":{"rendered":"https:\/\/nhdp.myds.me\/blog\/?p=204"},"modified":"2018-12-10T01:17:08","modified_gmt":"2018-12-09T18:17:08","slug":"tong-quan-don-gian-ve-mang-no-ron-tich-chap-convolutional-neural-networks","status":"publish","type":"post","link":"https:\/\/nhdp.myds.me\/blog\/2018\/11\/tong-quan-don-gian-ve-mang-no-ron-tich-chap-convolutional-neural-networks\/","title":{"rendered":"T\u1ed5ng quan \u0111\u01a1n gi\u1ea3n v\u1ec1 M\u1ea1ng n\u01a1-ron t\u00edch ch\u1eadp (Convolutional Neural Networks)"},"content":{"rendered":"

C\u1ea3m h\u1ee9ng c\u1ee7a M\u1ea1ng N\u01a1-ron T\u00edch Ch\u1eadp<\/h4>\n

T\u01b0\u01a1ng t\u1ef1 nh\u01b0 vi\u1ec7c tr\u1ebb em h\u1ecdc c\u00e1ch nh\u1eadn di\u1ec7n \u0111\u1ed1i t\u01b0\u1ee3ng, ch\u00fang ta c\u1ea7n cho thu\u1eadt to\u00e1n h\u1ecdc r\u1ea5t nhi\u1ec1u h\u00ecnh \u1ea3nh tr\u01b0\u1edbc khi n\u00f3 c\u00f3 th\u1ec3 \u0111\u01b0a ra d\u1ef1 \u0111o\u00e1n cho h\u00ecnh \u1ea3nh \u0111\u1ea7u v\u00e0o m\u00e0 n\u00f3 ch\u01b0a t\u1eebng th\u1ea5y.<\/p>\n

M\u00e1y t\u00ednh \u201cnh\u00ecn\u201d theo c\u00e1ch kh\u00e1c con ng\u01b0\u1eddi. Trong th\u1ebf gi\u1edbi m\u00e1y t\u00ednh ch\u1ec9 c\u00f3 nh\u1eefng con s\u1ed1. M\u1ed7i h\u00ecnh \u1ea3nh c\u00f3 th\u1ec3 \u0111\u01b0\u1ee3c bi\u1ec3u di\u1ec5n d\u01b0\u1edbi d\u1ea1ng m\u1ea3ng 2 chi\u1ec1u nh\u1eefng con s\u1ed1 \u0111\u01b0\u1ee3c g\u1ecdi l\u00e0 c\u00e1c pixel.<\/p>\n

M\u1eb7c d\u00f9 m\u00e1y t\u00ednh nh\u00ecn nh\u1eadn theo c\u00e1ch kh\u00e1c con ng\u01b0\u1eddi, ch\u00fang ta v\u1eabn c\u00f3 th\u1ec3 d\u1ea1y m\u00e1y t\u00ednh nh\u1eadn di\u1ec7n c\u00e1c m\u1eabu nh\u01b0 con ng\u01b0\u1eddi. \u0110i\u1ec1u quan tr\u1ecdng l\u00e0 ch\u00fang ta c\u1ea7n ngh\u0129 v\u1ec1 h\u00ecnh \u1ea3nh theo m\u1ed9t c\u00e1ch kh\u00e1c \u0111i.<\/p>\n

\"C\u00e1ch
C\u00e1ch m\u00e1y t\u00ednh “nh\u00ecn” m\u1ed9t h\u00ecnh. – Ngu\u1ed3n: http:\/\/cs231n.github.io\/classification\/<\/figcaption><\/figure>\n

<\/p>\n

\u0110\u1ec3 d\u1ea1y thu\u1eadt to\u00e1n nh\u1eadn di\u1ec7n \u0111\u1ed1i t\u01b0\u1ee3ng trong h\u00ecnh \u1ea3nh, ta s\u1eed d\u1ee5ng m\u1ed9t lo\u1ea1i M\u1ea1ng N\u01a1-ron Nh\u00e2n T\u1ea1o (Artificial Neural Network): M\u1ea1ng N\u01a1-ron T\u00edch Ch\u1eadp. T\u00ean c\u1ee7a n\u00f3 \u0111\u01b0\u1ee3c d\u1ef1a tr\u00ean ph\u00e9p t\u00ednh quan tr\u1ecdng \u0111\u01b0\u1ee3c s\u1eed d\u1ee5ng trong m\u1ea1ng \u0111\u00f3 l\u00e0 T\u00edch Ch\u1eadp.<\/p>\n

M\u1ea1ng N\u01a1-ron T\u00edch Ch\u1eadp l\u1ea5y c\u1ea3m h\u1ee9ng t\u1eeb n\u00e3o ng\u01b0\u1eddi. Nghi\u00ean c\u1ee9u trong nh\u1eefng th\u1eadp ni\u00ean 1950 v\u00e0 1960 c\u1ee7a D.H Hubel v\u00e0 T.N Wiesel tr\u00ean n\u00e3o c\u1ee7a \u0111\u1ed9ng v\u1eadt \u0111\u00e3 \u0111\u1ec1 xu\u1ea5t m\u1ed9t m\u00f4 h\u00ecnh m\u1edbi cho vi\u1ec7c c\u00e1ch m\u00e0 \u0111\u1ed9ng v\u1eadt nh\u00ecn nh\u1eadn th\u1ebf gi\u1edbi. Trong b\u00e1o c\u00e1o, hai \u00f4ng \u0111\u00e3 di\u1ec5n t\u1ea3 2 lo\u1ea1i t\u1ebf b\u00e0o n\u01a1-ron trong n\u00e3o v\u00e0 c\u00e1ch ho\u1ea1t \u0111\u1ed9ng kh\u00e1c nhau: t\u1ebf b\u00e0o \u0111\u01a1n gi\u1ea3n (simple cell \u2013 S cell) v\u00e0 t\u1ebf b\u00e0o ph\u1ee9c t\u1ea1p (complex cell \u2013 C cell).<\/p>\n

C\u00e1c t\u1ebf b\u00e0o \u0111\u01a1n gi\u1ea3n \u0111\u01b0\u1ee3c k\u00edch ho\u1ea1t khi nh\u1eadn di\u1ec7n c\u00e1c h\u00ecnh d\u00e1ng \u0111\u01a1n d\u1ea3n nh\u01b0 \u0111\u01b0\u1eddng n\u1eb1m trong m\u1ed9t khu v\u1ef1c c\u1ed1 \u0111\u1ecbnh v\u00e0 m\u1ed9t g\u00f3c c\u1ea1nh c\u1ee7a n\u00f3. C\u00e1c t\u1ebf b\u00e0o ph\u1ee9c t\u1ea1p c\u00f3 v\u00f9ng ti\u1ebfp nh\u1eadn l\u01a1n h\u01a1n v\u00e0 \u0111\u1ea7u ra c\u1ee7a n\u00f3 kh\u00f4ng nh\u1ea1y c\u1ea3m v\u1edbi nh\u1eefng v\u1ecb tr\u00ed c\u1ed1 \u0111\u1ecbnh trong v\u00f9ng.<\/p>\n

Trong th\u1ecb gi\u00e1c, v\u00f9ng ti\u1ebfp nh\u1eadn c\u1ee7a m\u1ed9t n\u01a1-ron t\u01b0\u01a1ng \u1ee9ng v\u1edbi m\u1ed9t v\u00f9ng tr\u00ean v\u00f5ng m\u1ea1c n\u01a1i m\u00e0 s\u1ebd k\u00edch ho\u1ea1t n\u01a1-ron t\u01b0\u01a1ng \u1ee9ng.<\/p>\n

N\u0103m 1980, Fukushima \u0111\u1ec1 xu\u1ea5t m\u00f4 h\u00ecnh m\u1ea1ng n\u01a1-ron c\u00f3 c\u1ea5p b\u1eadc g\u1ecdi l\u00e0 neocognitron<\/strong>. M\u00f4 h\u00ecnh n\u00e0y d\u1ef1a tr\u00ean kh\u00e1i ni\u1ec7m v\u1ec1 S cell v\u00e0 C cell. M\u1ea1n neocognitron c\u00f3 th\u1ec3 nh\u1eadn di\u1ec7n m\u1eabu d\u1ef1a tr\u00ean vi\u1ec7c h\u1ecdc h\u00ecnh d\u00e1ng c\u1ee7a \u0111\u1ed1i t\u01b0\u1ee3ng.<\/p>\n

Sau \u0111\u00f3 v\u00e0o n\u0103m 1998, M\u1ea1ng N\u01a1-ron T\u00edch Ch\u1eadp \u0111\u01b0\u1ee3c gi\u1edbi thi\u1ec7u b\u1edfi Bengio, Le Cun, Bottou v\u00e0 Haffner. M\u00f4 h\u00ecnh \u0111\u1ea7u ti\u00ean c\u1ee7a h\u1ecd \u0111\u01b0\u1ee3c g\u1ecdi t\u00ean l\u00e0 LeNet-5<\/strong>. M\u00f4 h\u00ecnh n\u00e0y c\u00f3 th\u1ec3 nh\u1eadn di\u1ec7n ch\u1eef s\u1ed1 vi\u1ebft tay.<\/p>\n

Ki\u1ebfn tr\u00fac M\u1ea1ng N\u01a1-ron T\u00edch Ch\u1eadp<\/h4>\n

M\u1ea1ng N\u01a1-ron T\u00edch Ch\u1eadp c\u00f3 ki\u1ebfn tr\u00fac kh\u00e1c v\u1edbi M\u1ea1ng N\u01a1-ron th\u00f4ng th\u01b0\u1eddng. M\u1ea1ng N\u01a1-ron b\u00ecnh th\u01b0\u1eddng chuy\u1ec3n \u0111\u1ed5i \u0111\u1ea7u v\u00e0o th\u00f4ng qua h\u00e0ng lo\u1ea1t c\u00e1c t\u1ea7ng \u1ea9n. M\u1ed7i t\u1ea7ng l\u00e0 m\u1ed9t t\u1eadp c\u00e1c n\u01a1-ron v\u00e0 c\u00e1c t\u1ea7ng \u0111\u01b0\u1ee3c li\u00ean k\u1ebft \u0111\u1ea7y \u0111\u1ee7 v\u1edbi c\u00e1c n\u01a1-ron \u1edf t\u1ea7ng tr\u01b0\u1edbc \u0111\u00f3. V\u00e0 \u1edf t\u1ea7ng cu\u1ed1i c\u00f9ng s\u1ebd l\u00e0 t\u1ea7ng k\u1ebft qu\u1ea3 \u0111\u1ea1i di\u1ec7n cho d\u1ef1 \u0111o\u00e1n c\u1ee7a m\u1ea1ng.<\/p>\n

\u0110\u1ea7u ti\u00ean, m\u1ea1ng N\u01a1-ron T\u00edch Ch\u1eadp \u0111\u01b0\u1ee3c chia th\u00e0nh 3 chi\u1ec1u: r\u1ed9ng, cao, v\u00e0 s\u00e2u. K\u1ebf \u0111\u00ean, c\u00e1c n\u01a1 ron trong m\u1ea1ng kh\u00f4ng li\u00ean k\u1ebft ho\u00e0n to\u00e0n v\u1edbi to\u00e0n b\u1ed9 n\u01a1-ron k\u1ebf \u0111\u1ebfn nh\u01b0ng ch\u1ec9 li\u00ean k\u1ebft t\u1edbi m\u1ed9t v\u00f9ng nh\u1ecf. Cu\u1ed1i c\u00f9ng, m\u1ed9t t\u1ea7ng \u0111\u1ea7u ra \u0111\u01b0\u1ee3c t\u1ed1i gi\u1ea3n th\u00e0nh v\u00e9c-t\u01a1 c\u1ee7a gi\u00e1 tr\u1ecb x\u00e1c su\u1ea5t.<\/p>\n

\"M\u1ea1ng
M\u1ea1ng n\u01a1-ron th\u00f4ng th\u01b0\u1eddng (tr\u00e1i) v\u00e0 CNN (ph\u1ea3i). – Ngu\u1ed3n: http:\/\/cs231n.github.io\/convolutional-networks\/<\/figcaption><\/figure>\n

CNNs g\u1ed3m hai th\u00e0nh ph\u1ea7n:<\/p>\n

    \n
  • Ph\u1ea7n t\u1ea7ng \u1ea9n hay ph\u1ea7n r\u00fat tr\u00edch \u0111\u1eb7c tr\u01b0ng: trong ph\u1ea7n n\u00e0y, m\u1ea1ng s\u1ebd ti\u1ebfn h\u00e0nh t\u00ednh to\u00e1n h\u00e0ng lo\u1ea1t ph\u00e9p t\u00edch ch\u1eadp<\/strong> v\u00e0 ph\u00e9p h\u1ee3p nh\u1ea5t<\/strong> (pooling) \u0111\u1ec3 ph\u00e1t hi\u1ec7n c\u00e1c \u0111\u1eb7c tr\u01b0ng. V\u00ed d\u1ee5: n\u1ebfu ta c\u00f3 h\u00ecnh \u1ea3nh con ng\u1ef1a v\u1eb1n, th\u00ec trong ph\u1ea7n n\u00e0y m\u1ea1ng s\u1ebd nh\u1eadn di\u1ec7n c\u00e1c s\u1ecdc v\u1eb1n, hai tai, v\u00e0 b\u1ed1n ch\u00e2n c\u1ee7a n\u00f3.<\/li>\n
  • Ph\u1ea7n ph\u00e2n l\u1edbp: t\u1ea1i ph\u1ea7n n\u00e0y, m\u1ed9t l\u1edbp v\u1edbi c\u00e1c li\u00ean k\u1ebft \u0111\u1ea7y \u0111\u1ee7 s\u1ebd \u0111\u00f3ng vai tr\u00f2 nh\u01b0 m\u1ed9t b\u1ed9 ph\u00e2n l\u1edbp c\u00e1c \u0111\u1eb7c tr\u01b0ng \u0111\u00e3 r\u00fat tr\u00edch \u0111\u01b0\u1ee3c tr\u01b0\u1edbc \u0111\u00f3. T\u1ea7ng n\u00e0y s\u1ebd \u0111\u01b0a ra x\u00e1c su\u1ea5t c\u1ee7a m\u1ed9t \u0111\u1ed1i t\u01b0\u1ee3ng trong h\u00ecnh.<\/li>\n<\/ul>\n
    \"Ki\u1ebfn
    Ki\u1ebfn tr\u00fac c\u1ee7a CNN. – Ngu\u1ed3n: https:\/\/www.mathworks.com\/videos\/introduction-to-deep-learning-what-are-convolutional-neural-networks–1489512765771.html<\/figcaption><\/figure>\n
    R\u00fat tr\u00edch \u0111\u1eb7c tr\u01b0ng<\/h5>\n

    T\u00edch ch\u1eadp l\u00e0 m\u1ed9t kh\u1ed1i quan tr\u1ecdng trong CNN. Thu\u1eadt ng\u1eef t\u00edch ch\u1eadp \u0111\u01b0\u1ee3c d\u1ef1a tr\u00ean m\u1ed9t ph\u00e9p h\u1ee3p nh\u1ea5t to\u00e1n h\u1ecdc c\u1ee7a hai h\u00e0m t\u1ea1o th\u00e0nh h\u00e0m th\u1ee9 ba. Ph\u00e9p to\u00e1n n\u00e0y k\u1ebft h\u1ee3p hai t\u1eadp th\u00f4ng tin kh\u00e1c nhau.<\/p>\n

    Trong tr\u01b0\u1eddng h\u1ee3p CNN, t\u00edch ch\u1eadp \u0111\u01b0\u1ee3c th\u1ef1c hi\u1ec7n tr\u00ean gi\u00e1 tr\u1ecb \u0111\u1ea7u v\u00e0o c\u1ee7a d\u1eef li\u1ec7u v\u00e0 kernel\/filter<\/strong> (thu\u1eadt ng\u1eef n\u00e0y \u0111\u01b0\u1ee3c s\u1eed d\u1ee5ng kh\u00e1c nhau t\u00f9y t\u00ecnh hu\u1ed1ng) \u0111\u1ec3 t\u1ea1o ra m\u1ed9t b\u1ea3n \u0111\u1ed3 \u0111\u1eb7c tr\u01b0ng<\/strong> (feature map).<\/p>\n

    Ta th\u1ef1c hi\u1ec7n ph\u00e9p t\u00edch ch\u1eadp b\u1eb1ng c\u00e1ch tr\u01b0\u1ee3t kernel\/filter theo d\u1eef li\u1ec7u \u0111\u1ea7u v\u00e0o. T\u1ea1i m\u1ed7i v\u1ecb tr\u00ed, ta ti\u1ebfn h\u00e0nh ph\u00e9p nh\u00e2n ma tr\u1eadn v\u00e0 t\u00ednh t\u1ed5ng c\u00e1c gi\u00e1 tr\u1ecb \u0111\u1ec3 \u0111\u01b0a v\u00e0o b\u1ea3n \u0111\u1ed3 \u0111\u1eb7c tr\u01b0ng.<\/p>\n

    Trong h\u00ecnh d\u01b0\u1edbi \u0111\u00e2y, th\u00e0nh ph\u1ea7n kernel\/filter (m\u00e0u xanh l\u00e1) tr\u01b0\u1ee3t tr\u00ean \u0111\u1ea7u v\u00e0o (m\u00e0u xanh d\u01b0\u01a1ng) v\u00e0 k\u1ebft qu\u1ea3 \u0111\u01b0\u1ee3c tr\u1ea3 v\u1ec1 b\u1ea3n \u0111\u1ed3 \u0111\u1eb7c tr\u01b0ng (m\u00e0u \u0111\u1ecf). Kernel\/filter c\u00f3 k\u00edch th\u01b0\u1edbc l\u00e0 3×3 trong v\u00ed d\u1ee5 n\u00e0y.<\/p>\n

    \"Ph\u00e9p
    Ph\u00e9p t\u00edch ch\u1eadp. – Ngu\u1ed3n: https:\/\/towardsdatascience.com\/applied-deep-learning-part-4-convolutional-neural-networks-584bc134c1e2<\/figcaption><\/figure>\n

    Trong th\u1ef1c t\u1ebf, t\u00edch ch\u1eadp \u0111\u01b0\u1ee3c th\u1ef1c hi\u1ec7n hi\u1ec7n tr\u00ean kh\u00f4ng gian 3 chi\u1ec1u. V\u00ec m\u1ed7i h\u00ecnh \u1ea3nh \u0111\u01b0\u1ee3c bi\u1ec3u di\u1ec5n d\u01b0\u1edbi d\u1ea1ng 3 chi\u1ec1u: r\u1ed9ng, cao, v\u00e0 s\u00e2u. Chi\u1ec1u s\u00e2u \u1edf \u0111\u00e2y ch\u00ednh l\u00e0 gi\u00e1 tr\u1ecb m\u00e0u s\u1eafc c\u1ee7a h\u00ecnh (RGB).<\/p>\n

    Ta th\u1ef1c hi\u1ec7n ph\u00e9p t\u00edch ch\u1eadp tr\u00ean \u0111\u1ea7u v\u00e0o nhi\u1ec1u l\u1ea7n kh\u00e1c nhau. M\u1ed7i l\u1ea7n s\u1eed d\u1ee5ng m\u1ed9t kernel\/filter kh\u00e1c nhau. K\u1ebft qu\u1ea3 ta s\u1ebd thu \u0111\u01b0\u1ee3c nh\u1eefng b\u1ea3n \u0111\u1ed3 \u0111\u1eb7c tr\u01b0ng kh\u00e1c nhau. Cu\u1ed1i c\u00f9ng, ta k\u1ebft h\u1ee3p to\u00e0n b\u1ed9 b\u1ea3n \u0111\u1ed3 \u0111\u1eb7c tr\u01b0ng n\u00e0y th\u00e0nh k\u1ebft qu\u1ea3 cu\u1ed1i c\u00f9ng c\u1ee7a t\u1ea7ng t\u00edch ch\u1eadp.<\/p>\n

    T\u01b0\u01a1ng t\u1ef1 nh\u01b0 m\u1ea1ng n\u01a1-ron th\u00f4ng th\u01b0\u1eddng, ta s\u1eed d\u1ee5ng m\u1ed9t h\u00e0m k\u00edch ho\u1ea1t (activate function) \u0111\u1ec3 c\u00f3 \u0111\u1ea7u ra d\u01b0\u1edbi d\u1ea1ng phi tuy\u1ebfn. Trong tr\u01b0\u1eddng h\u1ee3p CNN, \u0111\u1ea7u ra c\u1ee7a ph\u00e9p t\u00edch ch\u1eadp s\u1ebd \u0111i qua h\u00e0m k\u00edch ho\u1ea1t n\u00e0o \u0111\u00f3 v\u00ed d\u1ee5 nh\u01b0 h\u00e0m ReLU<\/strong> (rectified linear units).<\/p>\n

    Trong qu\u00e1 tr\u00ecnh tr\u01b0\u1ee3t kernel\/filter tr\u00ean d\u1eef li\u1ec7u \u0111\u1ea7u v\u00e0o, ta s\u1ebd quy \u0111\u1ecbnh m\u1ed9t b\u01b0\u1edbc nh\u1ea3y (stride) v\u1edbi m\u1ed7i l\u1ea7n di chuy\u1ec3n. Th\u00f4ng th\u01b0\u1eddng ta l\u1ef1a ch\u1ecdn th\u01b0\u1eddng ch\u1ecdn b\u01b0\u1edbc nh\u1ea3y l\u00e0 1. N\u1ebfu k\u00edch th\u01b0\u1edbc b\u01b0\u1edbc nh\u1ea3y t\u0103ng, kernel\/filter s\u1ebd c\u00f3 \u00edt \u00f4 tr\u00f9ng l\u1eafp.<\/p>\n

    B\u1edfi v\u00ec k\u00edch th\u01b0\u1edbc \u0111\u1ea7u ra lu\u00f4n nh\u1ecf h\u01a1n \u0111\u1ea7u v\u00e0o n\u00ean ta c\u1ea7n m\u1ed9t ph\u00e9p x\u1eed l\u00ed \u0111\u1ea7u v\u00e0o \u0111\u1ec3 \u0111\u1ea7u ra kh\u00f4ng b\u1ecb co gi\u00e3n. \u0110\u01a1n gi\u1ea3n ta ch\u1ec9 c\u1ea7n th\u00eam m\u1ed9t l\u1ec1 nh\u1ecf v\u00e0o \u0111\u1ea7u v\u00e0o. M\u1ed9t l\u1ec1 v\u1edbi gi\u00e1 tr\u1ecb 0 s\u1ebd \u0111\u01b0\u1ee3c th\u00eam v\u00e0o xung quanh \u0111\u1ea7u v\u00e0o tr\u01b0\u1edbc khi th\u1ef1c hi\u1ec7n ph\u00e9p t\u00edch ch\u1eadp.<\/p>\n

    Th\u00f4ng th\u01b0\u1eddng, sau m\u1ed7i t\u1ea7ng t\u00edch ch\u1eadp, ta s\u1ebd cho k\u1ebft qu\u1ea3 \u0111i qua m\u1ed9t t\u1ea7ng h\u1ee3p nh\u1ea5t<\/strong> (pooling layer). M\u1ee5c \u0111\u00edch c\u1ee7a t\u1ea7ng n\u00e0y l\u00e0 \u0111\u1ec3 nhanh ch\u00f3ng gi\u1ea3m s\u1ed1 chi\u1ec1u. Vi\u1ec7c n\u00e0y gi\u00fap gi\u1ea3m th\u1eddi gian h\u1ecdc v\u00e0 h\u1ea1n ch\u1ebf vi\u1ec7c overfitting<\/strong>.<\/p>\n

    M\u1ed9t ph\u00e9p h\u1ee3p nh\u1ea5t \u0111\u01a1n gi\u1ea3n th\u01b0\u1eddng \u0111\u01b0\u1ee3c d\u00f9ng \u0111\u00f3 l\u00e0 max pooling<\/strong>, ph\u00e9p n\u00e0y l\u1ea5y gi\u00e1 tr\u1ecb l\u1edbn nh\u1ea5t c\u1ee7a m\u1ed9t v\u00f9ng \u0111\u1ec3 \u0111\u1ea1i di\u1ec7n cho v\u00f9ng \u0111\u00f3. K\u00edch th\u01b0\u1edbc c\u1ee7a v\u00f9ng s\u1ebd \u0111\u01b0\u1ee3c x\u00e1c \u0111\u1ecbnh tr\u01b0\u1edbc \u0111\u1ec3 gi\u1ea3m k\u00edch th\u01b0\u1edbc c\u1ee7a b\u1ea3n \u0111\u1ed3 \u0111\u1eb7c tr\u01b0ng nhanh ch\u00f3ng nh\u01b0ng v\u1eabn gi\u1eef \u0111\u01b0\u1ee3c th\u00f4ng tin c\u1ea7n thi\u1ebft.<\/p>\n

    \"Max
    Max pooling k\u00edch th\u01b0\u1edbc 2×2. – Ngu\u1ed3n:  http:\/\/cs231n.github.io\/convolutional-networks\/<\/figcaption><\/figure>\n

    T\u1ed5ng k\u1ebft l\u1ea1i khi s\u1eed d\u1ee5ng CNN, ta c\u1ea7n ch\u00fa \u00fd \u0111\u1ebfn 4 si\u00eau tham s\u1ed1 quan tr\u1ecdng:<\/p>\n

      \n
    • K\u00edch th\u01b0\u1edbc kernel\/filter<\/li>\n
    • S\u1ed1 l\u01b0\u1ee3ng kernel\/filter<\/li>\n
    • K\u00edch th\u01b0\u1edbc b\u01b0\u1edbc nh\u1ea3y (stride)<\/li>\n
    • K\u00edch th\u01b0\u1edbc l\u1ec1 (padding)<\/li>\n<\/ul>\n
      Ph\u00e2n l\u1edbp<\/h5>\n

      Trong ph\u1ea7n ph\u00e2n l\u1edbp, ta s\u1eed d\u1ee5ng m\u1ed9t v\u00e0i t\u1ea7ng v\u1edbi k\u1ebft n\u1ed1i \u0111\u1ea7y \u0111\u1ee7 \u0111\u1ec3 x\u1eed l\u00ed k\u1ebft qu\u1ea3 c\u1ee7a ph\u1ea7n t\u00edch ch\u1eadp. V\u00ec \u0111\u1ea7u v\u00e0o c\u1ee7a m\u1ea1ng li\u00ean k\u1ebft \u0111\u1ea7y \u0111\u1ee7 l\u00e0 1 chi\u1ec1u, ta c\u1ea7n l\u00e0m ph\u1eb3ng \u0111\u1ea7u v\u00e0o tr\u01b0\u1edbc khi ph\u00e2n l\u1edbp. T\u1ea7ng cu\u1ed1i c\u00f9ng trong m\u1ea1ng CNN l\u00e0 m\u1ed9t t\u1ea7ng li\u00ean k\u1ebft \u0111\u1ea7y \u0111\u1ee7, ph\u1ea7n n\u00e0y ho\u1ea1t \u0111\u1ed9ng t\u01b0\u01a1ng t\u1ef1 nh\u01b0 m\u1ea1ng n\u01a1-ron th\u00f4ng th\u01b0\u1eddng.<\/p>\n

      K\u1ebft qu\u1ea3 thu \u0111\u01b0\u1ee3c cu\u1ed1i c\u00f9ng c\u0169ng s\u1ebd l\u00e0 m\u1ed9t v\u00e9c-t\u01a1 v\u1edbi c\u00e1c gi\u00e1 tr\u1ecb x\u00e1c su\u1ea5t cho vi\u1ec7c d\u1ef1 \u0111o\u00e1n nh\u01b0 m\u1ea1ng n\u01a1-ron th\u00f4ng th\u01b0\u1eddng.<\/p>\n

      \"V\u00ed
      V\u00ed d\u1ee5 cho t\u00edch ch\u1eadp v\u1edbi 2 filter, 3 c\u1ed9t, 2 b\u01b0\u1edbc nh\u1ea3y, l\u1ec1 = 1. – Ngu\u1ed3n: http:\/\/cs231n.github.io\/convolutional-networks\/<\/figcaption><\/figure>\n

      Ngu\u1ed3n: https:\/\/medium.freecodecamp.org\/an-intuitive-guide-to-convolutional-neural-networks-260c2de0a050<\/p>\n","protected":false},"excerpt":{"rendered":"

      T\u01b0\u01a1ng t\u1ef1 nh\u01b0 vi\u1ec7c tr\u1ebb em h\u1ecdc c\u00e1ch nh\u1eadn di\u1ec7n \u0111\u1ed1i t\u01b0\u1ee3ng, ch\u00fang ta c\u1ea7n cho thu\u1eadt to\u00e1n h\u1ecdc r\u1ea5t nhi\u1ec1u h\u00ecnh \u1ea3nh tr\u01b0\u1edbc khi n\u00f3 c\u00f3 th\u1ec3 \u0111\u01b0a ra d\u1ef1 \u0111o\u00e1n cho h\u00ecnh \u1ea3nh \u0111\u1ea7u v\u00e0o m\u00e0 n\u00f3 ch\u01b0a t\u1eebng th\u1ea5y.<\/p>\n

      M\u00e1y t\u00ednh \u201cnh\u00ecn\u201d theo c\u00e1ch kh\u00e1c con ng\u01b0\u1eddi. Trong th\u1ebf gi\u1edbi m\u00e1y t\u00ednh ch\u1ec9 c\u00f3 nh\u1eefng con s\u1ed1. M\u1ed7i h\u00ecnh \u1ea3nh c\u00f3 th\u1ec3 \u0111\u01b0\u1ee3c bi\u1ec3u di\u1ec5n d\u01b0\u1edbi d\u1ea1ng m\u1ea3ng 2 chi\u1ec1u nh\u1eefng con s\u1ed1 \u0111\u01b0\u1ee3c g\u1ecdi l\u00e0 c\u00e1c pixel.<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_jetpack_newsletter_access":"","_jetpack_dont_email_post_to_subs":false,"_jetpack_newsletter_tier_id":0,"_jetpack_memberships_contains_paywalled_content":false,"_jetpack_memberships_contains_paid_content":false,"footnotes":"","jetpack_post_was_ever_published":false},"categories":[16],"tags":[18,17,19,22,21,20],"class_list":["post-204","post","type-post","status-publish","format-standard","hentry","category-khoa-hoc-may-tinh","tag-classification","tag-cnn","tag-computer-vision","tag-convolutional","tag-machine-learning","tag-neuron-network"],"yoast_head":"\nT\u1ed5ng quan \u0111\u01a1n gi\u1ea3n v\u1ec1 M\u1ea1ng n\u01a1-ron t\u00edch ch\u1eadp (Convolutional Neural Networks) - Soli Ph\u01b0\u01a1ng is Me<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/nhdp.myds.me\/blog\/2018\/11\/tong-quan-don-gian-ve-mang-no-ron-tich-chap-convolutional-neural-networks\/\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"T\u1ed5ng quan \u0111\u01a1n gi\u1ea3n v\u1ec1 M\u1ea1ng n\u01a1-ron t\u00edch ch\u1eadp (Convolutional Neural Networks) - Soli Ph\u01b0\u01a1ng is Me\" \/>\n<meta property=\"og:description\" content=\"T\u01b0\u01a1ng t\u1ef1 nh\u01b0 vi\u1ec7c tr\u1ebb em h\u1ecdc c\u00e1ch nh\u1eadn di\u1ec7n \u0111\u1ed1i t\u01b0\u1ee3ng, ch\u00fang ta c\u1ea7n cho thu\u1eadt to\u00e1n h\u1ecdc r\u1ea5t nhi\u1ec1u h\u00ecnh \u1ea3nh tr\u01b0\u1edbc khi n\u00f3 c\u00f3 th\u1ec3 \u0111\u01b0a ra d\u1ef1 \u0111o\u00e1n cho h\u00ecnh \u1ea3nh \u0111\u1ea7u v\u00e0o m\u00e0 n\u00f3 ch\u01b0a t\u1eebng th\u1ea5y. M\u00e1y t\u00ednh \u201cnh\u00ecn\u201d theo c\u00e1ch kh\u00e1c con ng\u01b0\u1eddi. Trong th\u1ebf gi\u1edbi m\u00e1y t\u00ednh ch\u1ec9 c\u00f3 nh\u1eefng con s\u1ed1. M\u1ed7i h\u00ecnh \u1ea3nh c\u00f3 th\u1ec3 \u0111\u01b0\u1ee3c bi\u1ec3u di\u1ec5n d\u01b0\u1edbi d\u1ea1ng m\u1ea3ng 2 chi\u1ec1u nh\u1eefng con s\u1ed1 \u0111\u01b0\u1ee3c g\u1ecdi l\u00e0 c\u00e1c pixel.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/nhdp.myds.me\/blog\/2018\/11\/tong-quan-don-gian-ve-mang-no-ron-tich-chap-convolutional-neural-networks\/\" \/>\n<meta property=\"og:site_name\" content=\"Soli Ph\u01b0\u01a1ng is Me\" \/>\n<meta property=\"article:publisher\" content=\"https:\/\/www.facebook.com\/nhdphuong93\" \/>\n<meta property=\"article:author\" content=\"https:\/\/www.facebook.com\/nhdphuong93\" \/>\n<meta property=\"article:published_time\" content=\"2018-11-24T05:53:15+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2018-12-09T18:17:08+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/nhdp.myds.me\/blog\/wp-content\/uploads\/2018\/11\/1.png\" \/>\n<meta name=\"author\" content=\"Soli Ph\u01b0\u01a1ng\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:creator\" content=\"@nhdphuong93\" \/>\n<meta name=\"twitter:site\" content=\"@nhdphuong93\" \/>\n<meta name=\"twitter:label1\" content=\"Written by\" \/>\n\t<meta name=\"twitter:data1\" content=\"Soli Ph\u01b0\u01a1ng\" \/>\n\t<meta name=\"twitter:label2\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data2\" content=\"11 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\\\/\\\/nhdp.myds.me\\\/blog\\\/2018\\\/11\\\/tong-quan-don-gian-ve-mang-no-ron-tich-chap-convolutional-neural-networks\\\/#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/nhdp.myds.me\\\/blog\\\/2018\\\/11\\\/tong-quan-don-gian-ve-mang-no-ron-tich-chap-convolutional-neural-networks\\\/\"},\"author\":{\"name\":\"Soli Ph\u01b0\u01a1ng\",\"@id\":\"https:\\\/\\\/nhdp.myds.me\\\/blog\\\/#\\\/schema\\\/person\\\/1b81c25c72af317e7807816fcd74ede0\"},\"headline\":\"T\u1ed5ng quan \u0111\u01a1n gi\u1ea3n v\u1ec1 M\u1ea1ng n\u01a1-ron t\u00edch ch\u1eadp (Convolutional Neural Networks)\",\"datePublished\":\"2018-11-24T05:53:15+00:00\",\"dateModified\":\"2018-12-09T18:17:08+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/nhdp.myds.me\\\/blog\\\/2018\\\/11\\\/tong-quan-don-gian-ve-mang-no-ron-tich-chap-convolutional-neural-networks\\\/\"},\"wordCount\":2253,\"commentCount\":0,\"publisher\":{\"@id\":\"https:\\\/\\\/nhdp.myds.me\\\/blog\\\/#\\\/schema\\\/person\\\/1b81c25c72af317e7807816fcd74ede0\"},\"image\":{\"@id\":\"https:\\\/\\\/nhdp.myds.me\\\/blog\\\/2018\\\/11\\\/tong-quan-don-gian-ve-mang-no-ron-tich-chap-convolutional-neural-networks\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/nhdp.myds.me\\\/blog\\\/wp-content\\\/uploads\\\/2018\\\/11\\\/1.png\",\"keywords\":[\"classification\",\"cnn\",\"computer vision\",\"convolutional\",\"machine learning\",\"neuron network\"],\"articleSection\":[\"Khoa h\u1ecdc M\u00e1y t\u00ednh\"],\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"CommentAction\",\"name\":\"Comment\",\"target\":[\"https:\\\/\\\/nhdp.myds.me\\\/blog\\\/2018\\\/11\\\/tong-quan-don-gian-ve-mang-no-ron-tich-chap-convolutional-neural-networks\\\/#respond\"]}]},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/nhdp.myds.me\\\/blog\\\/2018\\\/11\\\/tong-quan-don-gian-ve-mang-no-ron-tich-chap-convolutional-neural-networks\\\/\",\"url\":\"https:\\\/\\\/nhdp.myds.me\\\/blog\\\/2018\\\/11\\\/tong-quan-don-gian-ve-mang-no-ron-tich-chap-convolutional-neural-networks\\\/\",\"name\":\"T\u1ed5ng quan \u0111\u01a1n gi\u1ea3n v\u1ec1 M\u1ea1ng n\u01a1-ron t\u00edch ch\u1eadp (Convolutional Neural Networks) - Soli Ph\u01b0\u01a1ng is Me\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/nhdp.myds.me\\\/blog\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/nhdp.myds.me\\\/blog\\\/2018\\\/11\\\/tong-quan-don-gian-ve-mang-no-ron-tich-chap-convolutional-neural-networks\\\/#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/nhdp.myds.me\\\/blog\\\/2018\\\/11\\\/tong-quan-don-gian-ve-mang-no-ron-tich-chap-convolutional-neural-networks\\\/#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/nhdp.myds.me\\\/blog\\\/wp-content\\\/uploads\\\/2018\\\/11\\\/1.png\",\"datePublished\":\"2018-11-24T05:53:15+00:00\",\"dateModified\":\"2018-12-09T18:17:08+00:00\",\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/nhdp.myds.me\\\/blog\\\/2018\\\/11\\\/tong-quan-don-gian-ve-mang-no-ron-tich-chap-convolutional-neural-networks\\\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/nhdp.myds.me\\\/blog\\\/2018\\\/11\\\/tong-quan-don-gian-ve-mang-no-ron-tich-chap-convolutional-neural-networks\\\/#primaryimage\",\"url\":\"https:\\\/\\\/i0.wp.com\\\/nhdp.myds.me\\\/blog\\\/wp-content\\\/uploads\\\/2018\\\/11\\\/1.png?fit=600%2C346&ssl=1\",\"contentUrl\":\"https:\\\/\\\/i0.wp.com\\\/nhdp.myds.me\\\/blog\\\/wp-content\\\/uploads\\\/2018\\\/11\\\/1.png?fit=600%2C346&ssl=1\",\"width\":600,\"height\":346,\"caption\":\"C\u00e1ch m\u00e1y t\u00ednh \\\"nh\u00ecn\\\" m\u1ed9t h\u00ecnh. - Ngu\u1ed3n: http:\\\/\\\/cs231n.github.io\\\/classification\\\/\"},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/nhdp.myds.me\\\/blog\\\/#website\",\"url\":\"https:\\\/\\\/nhdp.myds.me\\\/blog\\\/\",\"name\":\"Soli Ph\u01b0\u01a1ng is Me\",\"description\":\"G\u00f3c nh\u1ecf gi\u1eefa \u0111\u1ea1i d\u01b0\u01a1ng Internet c\u1ee7a Soli\",\"publisher\":{\"@id\":\"https:\\\/\\\/nhdp.myds.me\\\/blog\\\/#\\\/schema\\\/person\\\/1b81c25c72af317e7807816fcd74ede0\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/nhdp.myds.me\\\/blog\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"},{\"@type\":[\"Person\",\"Organization\"],\"@id\":\"https:\\\/\\\/nhdp.myds.me\\\/blog\\\/#\\\/schema\\\/person\\\/1b81c25c72af317e7807816fcd74ede0\",\"name\":\"Soli Ph\u01b0\u01a1ng\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/i0.wp.com\\\/nhdp.myds.me\\\/blog\\\/wp-content\\\/uploads\\\/2019\\\/11\\\/0022_15A.jpeg?fit=2048%2C1358&ssl=1\",\"url\":\"https:\\\/\\\/i0.wp.com\\\/nhdp.myds.me\\\/blog\\\/wp-content\\\/uploads\\\/2019\\\/11\\\/0022_15A.jpeg?fit=2048%2C1358&ssl=1\",\"contentUrl\":\"https:\\\/\\\/i0.wp.com\\\/nhdp.myds.me\\\/blog\\\/wp-content\\\/uploads\\\/2019\\\/11\\\/0022_15A.jpeg?fit=2048%2C1358&ssl=1\",\"width\":2048,\"height\":1358,\"caption\":\"Soli Ph\u01b0\u01a1ng\"},\"logo\":{\"@id\":\"https:\\\/\\\/i0.wp.com\\\/nhdp.myds.me\\\/blog\\\/wp-content\\\/uploads\\\/2019\\\/11\\\/0022_15A.jpeg?fit=2048%2C1358&ssl=1\"},\"sameAs\":[\"https:\\\/\\\/nhdp.myds.me\\\/\",\"https:\\\/\\\/www.facebook.com\\\/nhdphuong93\",\"https:\\\/\\\/x.com\\\/nhdphuong93\"]}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"T\u1ed5ng quan \u0111\u01a1n gi\u1ea3n v\u1ec1 M\u1ea1ng n\u01a1-ron t\u00edch ch\u1eadp (Convolutional Neural Networks) - Soli Ph\u01b0\u01a1ng is Me","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/nhdp.myds.me\/blog\/2018\/11\/tong-quan-don-gian-ve-mang-no-ron-tich-chap-convolutional-neural-networks\/","og_locale":"en_US","og_type":"article","og_title":"T\u1ed5ng quan \u0111\u01a1n gi\u1ea3n v\u1ec1 M\u1ea1ng n\u01a1-ron t\u00edch ch\u1eadp (Convolutional Neural Networks) - Soli Ph\u01b0\u01a1ng is Me","og_description":"T\u01b0\u01a1ng t\u1ef1 nh\u01b0 vi\u1ec7c tr\u1ebb em h\u1ecdc c\u00e1ch nh\u1eadn di\u1ec7n \u0111\u1ed1i t\u01b0\u1ee3ng, ch\u00fang ta c\u1ea7n cho thu\u1eadt to\u00e1n h\u1ecdc r\u1ea5t nhi\u1ec1u h\u00ecnh \u1ea3nh tr\u01b0\u1edbc khi n\u00f3 c\u00f3 th\u1ec3 \u0111\u01b0a ra d\u1ef1 \u0111o\u00e1n cho h\u00ecnh \u1ea3nh \u0111\u1ea7u v\u00e0o m\u00e0 n\u00f3 ch\u01b0a t\u1eebng th\u1ea5y. M\u00e1y t\u00ednh \u201cnh\u00ecn\u201d theo c\u00e1ch kh\u00e1c con ng\u01b0\u1eddi. Trong th\u1ebf gi\u1edbi m\u00e1y t\u00ednh ch\u1ec9 c\u00f3 nh\u1eefng con s\u1ed1. M\u1ed7i h\u00ecnh \u1ea3nh c\u00f3 th\u1ec3 \u0111\u01b0\u1ee3c bi\u1ec3u di\u1ec5n d\u01b0\u1edbi d\u1ea1ng m\u1ea3ng 2 chi\u1ec1u nh\u1eefng con s\u1ed1 \u0111\u01b0\u1ee3c g\u1ecdi l\u00e0 c\u00e1c pixel.","og_url":"https:\/\/nhdp.myds.me\/blog\/2018\/11\/tong-quan-don-gian-ve-mang-no-ron-tich-chap-convolutional-neural-networks\/","og_site_name":"Soli Ph\u01b0\u01a1ng is Me","article_publisher":"https:\/\/www.facebook.com\/nhdphuong93","article_author":"https:\/\/www.facebook.com\/nhdphuong93","article_published_time":"2018-11-24T05:53:15+00:00","article_modified_time":"2018-12-09T18:17:08+00:00","og_image":[{"url":"https:\/\/nhdp.myds.me\/blog\/wp-content\/uploads\/2018\/11\/1.png","type":"","width":"","height":""}],"author":"Soli Ph\u01b0\u01a1ng","twitter_card":"summary_large_image","twitter_creator":"@nhdphuong93","twitter_site":"@nhdphuong93","twitter_misc":{"Written by":"Soli Ph\u01b0\u01a1ng","Est. reading time":"11 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/nhdp.myds.me\/blog\/2018\/11\/tong-quan-don-gian-ve-mang-no-ron-tich-chap-convolutional-neural-networks\/#article","isPartOf":{"@id":"https:\/\/nhdp.myds.me\/blog\/2018\/11\/tong-quan-don-gian-ve-mang-no-ron-tich-chap-convolutional-neural-networks\/"},"author":{"name":"Soli Ph\u01b0\u01a1ng","@id":"https:\/\/nhdp.myds.me\/blog\/#\/schema\/person\/1b81c25c72af317e7807816fcd74ede0"},"headline":"T\u1ed5ng quan \u0111\u01a1n gi\u1ea3n v\u1ec1 M\u1ea1ng n\u01a1-ron t\u00edch ch\u1eadp (Convolutional Neural Networks)","datePublished":"2018-11-24T05:53:15+00:00","dateModified":"2018-12-09T18:17:08+00:00","mainEntityOfPage":{"@id":"https:\/\/nhdp.myds.me\/blog\/2018\/11\/tong-quan-don-gian-ve-mang-no-ron-tich-chap-convolutional-neural-networks\/"},"wordCount":2253,"commentCount":0,"publisher":{"@id":"https:\/\/nhdp.myds.me\/blog\/#\/schema\/person\/1b81c25c72af317e7807816fcd74ede0"},"image":{"@id":"https:\/\/nhdp.myds.me\/blog\/2018\/11\/tong-quan-don-gian-ve-mang-no-ron-tich-chap-convolutional-neural-networks\/#primaryimage"},"thumbnailUrl":"https:\/\/nhdp.myds.me\/blog\/wp-content\/uploads\/2018\/11\/1.png","keywords":["classification","cnn","computer vision","convolutional","machine learning","neuron network"],"articleSection":["Khoa h\u1ecdc M\u00e1y t\u00ednh"],"inLanguage":"en-US","potentialAction":[{"@type":"CommentAction","name":"Comment","target":["https:\/\/nhdp.myds.me\/blog\/2018\/11\/tong-quan-don-gian-ve-mang-no-ron-tich-chap-convolutional-neural-networks\/#respond"]}]},{"@type":"WebPage","@id":"https:\/\/nhdp.myds.me\/blog\/2018\/11\/tong-quan-don-gian-ve-mang-no-ron-tich-chap-convolutional-neural-networks\/","url":"https:\/\/nhdp.myds.me\/blog\/2018\/11\/tong-quan-don-gian-ve-mang-no-ron-tich-chap-convolutional-neural-networks\/","name":"T\u1ed5ng quan \u0111\u01a1n gi\u1ea3n v\u1ec1 M\u1ea1ng n\u01a1-ron t\u00edch ch\u1eadp (Convolutional Neural Networks) - Soli Ph\u01b0\u01a1ng is Me","isPartOf":{"@id":"https:\/\/nhdp.myds.me\/blog\/#website"},"primaryImageOfPage":{"@id":"https:\/\/nhdp.myds.me\/blog\/2018\/11\/tong-quan-don-gian-ve-mang-no-ron-tich-chap-convolutional-neural-networks\/#primaryimage"},"image":{"@id":"https:\/\/nhdp.myds.me\/blog\/2018\/11\/tong-quan-don-gian-ve-mang-no-ron-tich-chap-convolutional-neural-networks\/#primaryimage"},"thumbnailUrl":"https:\/\/nhdp.myds.me\/blog\/wp-content\/uploads\/2018\/11\/1.png","datePublished":"2018-11-24T05:53:15+00:00","dateModified":"2018-12-09T18:17:08+00:00","inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/nhdp.myds.me\/blog\/2018\/11\/tong-quan-don-gian-ve-mang-no-ron-tich-chap-convolutional-neural-networks\/"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/nhdp.myds.me\/blog\/2018\/11\/tong-quan-don-gian-ve-mang-no-ron-tich-chap-convolutional-neural-networks\/#primaryimage","url":"https:\/\/i0.wp.com\/nhdp.myds.me\/blog\/wp-content\/uploads\/2018\/11\/1.png?fit=600%2C346&ssl=1","contentUrl":"https:\/\/i0.wp.com\/nhdp.myds.me\/blog\/wp-content\/uploads\/2018\/11\/1.png?fit=600%2C346&ssl=1","width":600,"height":346,"caption":"C\u00e1ch m\u00e1y t\u00ednh \"nh\u00ecn\" m\u1ed9t h\u00ecnh. - Ngu\u1ed3n: http:\/\/cs231n.github.io\/classification\/"},{"@type":"WebSite","@id":"https:\/\/nhdp.myds.me\/blog\/#website","url":"https:\/\/nhdp.myds.me\/blog\/","name":"Soli Ph\u01b0\u01a1ng is Me","description":"G\u00f3c nh\u1ecf gi\u1eefa \u0111\u1ea1i d\u01b0\u01a1ng Internet c\u1ee7a Soli","publisher":{"@id":"https:\/\/nhdp.myds.me\/blog\/#\/schema\/person\/1b81c25c72af317e7807816fcd74ede0"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/nhdp.myds.me\/blog\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"},{"@type":["Person","Organization"],"@id":"https:\/\/nhdp.myds.me\/blog\/#\/schema\/person\/1b81c25c72af317e7807816fcd74ede0","name":"Soli Ph\u01b0\u01a1ng","image":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/i0.wp.com\/nhdp.myds.me\/blog\/wp-content\/uploads\/2019\/11\/0022_15A.jpeg?fit=2048%2C1358&ssl=1","url":"https:\/\/i0.wp.com\/nhdp.myds.me\/blog\/wp-content\/uploads\/2019\/11\/0022_15A.jpeg?fit=2048%2C1358&ssl=1","contentUrl":"https:\/\/i0.wp.com\/nhdp.myds.me\/blog\/wp-content\/uploads\/2019\/11\/0022_15A.jpeg?fit=2048%2C1358&ssl=1","width":2048,"height":1358,"caption":"Soli Ph\u01b0\u01a1ng"},"logo":{"@id":"https:\/\/i0.wp.com\/nhdp.myds.me\/blog\/wp-content\/uploads\/2019\/11\/0022_15A.jpeg?fit=2048%2C1358&ssl=1"},"sameAs":["https:\/\/nhdp.myds.me\/","https:\/\/www.facebook.com\/nhdphuong93","https:\/\/x.com\/nhdphuong93"]}]}},"jetpack_featured_media_url":"","jetpack_sharing_enabled":true,"jetpack_shortlink":"https:\/\/wp.me\/phcIBA-3i","jetpack_likes_enabled":true,"amp_enabled":true,"_links":{"self":[{"href":"https:\/\/nhdp.myds.me\/blog\/wp-json\/wp\/v2\/posts\/204","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/nhdp.myds.me\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/nhdp.myds.me\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/nhdp.myds.me\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/nhdp.myds.me\/blog\/wp-json\/wp\/v2\/comments?post=204"}],"version-history":[{"count":0,"href":"https:\/\/nhdp.myds.me\/blog\/wp-json\/wp\/v2\/posts\/204\/revisions"}],"wp:attachment":[{"href":"https:\/\/nhdp.myds.me\/blog\/wp-json\/wp\/v2\/media?parent=204"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/nhdp.myds.me\/blog\/wp-json\/wp\/v2\/categories?post=204"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/nhdp.myds.me\/blog\/wp-json\/wp\/v2\/tags?post=204"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}