Tensorflow CNN turorial -...
Transcript of Tensorflow CNN turorial -...
![Page 1: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/1.jpg)
Tensorflow CNNturorial2017/03/10
![Page 2: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/2.jpg)
Lenet-5
[LeCun etal.,1998]
![Page 3: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/3.jpg)
Today’sexample
Imagesource
![Page 4: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/4.jpg)
![Page 5: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/5.jpg)
32
3
Convolution Layer
32x32x3 image
width
height
32
depth
![Page 6: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/6.jpg)
32
32
3
Convolution Layer
5x5x3 filter
32x32x3 image
Convolve the filter with the imagei.e. “slide over the image spatially, computing dot products”
![Page 7: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/7.jpg)
32
32
3
Convolution Layer
5x5x3 filter
32x32x3 image
Convolve the filter with the imagei.e. “slide over the image spatially, computing dot products”
Filters always extend the full depth of the input volume
![Page 8: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/8.jpg)
32
32
3
Convolution Layer
32x32x3 image 5x5x3 filter
1 number:the result of taking a dot product between the filter and a small 5x5x3 chunk of the image(i.e. 5*5*3 = 75-dimensional dot product + bias)
![Page 9: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/9.jpg)
7
7x7 input (spatially) assume 3x3 filter
7
![Page 10: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/10.jpg)
7
7x7 input (spatially) assume 3x3 filter
7
![Page 11: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/11.jpg)
7
7x7 input (spatially) assume 3x3 filter
7
![Page 12: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/12.jpg)
7
7x7 input (spatially) assume 3x3 filter
7
![Page 13: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/13.jpg)
=> 5x5 output
7
7x7 input (spatially) assume 3x3 filter
7
![Page 14: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/14.jpg)
32
32
3
Convolution Layeractivation map
32x32x3 image5x5x3 filter
1
28
28
convolve (slide) over all spatial locations
![Page 15: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/15.jpg)
32
32
3
Convolution Layer
32x32x3 image 5x5x3 filter
activation maps
1
28
28
convolve (slide) over all spatial locations
consider a second, green filter
![Page 16: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/16.jpg)
32
3 6
28
activation maps
32
28
Convolution Layer
For example, if we had 6 5x5 filters, we’ll get 6 separate activation maps:
We stack these up to get a “new image” of size 28x28x6!
![Page 17: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/17.jpg)
Preview: ConvNet is a sequence of Convolution Layers, interspersed with activation functions
32
32
3
28
28
6
CONV, ReLUe.g. 65x5x3filters
![Page 18: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/18.jpg)
Preview: ConvNet is a sequence of Convolutional Layers, interspersed with activation functions
32
32
3
CONV, ReLUe.g. 65x5x3filters 28
28
6
CONV, ReLUe.g. 10 5x5x6 filters
CONV, ReLU
….
10
24
24
![Page 19: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/19.jpg)
32x32 input convolved repeatedly with 5x5 filters shrinks volumes spatially! (32 -> 28 -> 24 ...). Shrinking too fast is not good, doesn’t work well.
32
32
3
CONV, ReLUe.g. 65x5x3filters 28
28
6
CONV, ReLUe.g. 10 5x5x6 filters
CONV, ReLU
….
10
24
24
![Page 20: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/20.jpg)
7
7x7 input (spatially) assume 3x3 filter
7
A closer look at spatial dimensions:
![Page 21: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/21.jpg)
7
7x7 input (spatially) assume 3x3 filter
7
A closer look at spatial dimensions:
![Page 22: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/22.jpg)
7
7x7 input (spatially) assume 3x3 filter
7
A closer look at spatial dimensions:
![Page 23: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/23.jpg)
7
7x7 input (spatially) assume 3x3 filter
7
A closer look at spatial dimensions:
![Page 24: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/24.jpg)
=> 5x5 output
7
7x7 input (spatially) assume 3x3 filter
7
A closer look at spatial dimensions:
![Page 25: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/25.jpg)
7x7 input (spatially) assume 3x3 filter applied with stride 2
7
7
A closer look at spatial dimensions:
![Page 26: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/26.jpg)
7x7 input (spatially) assume 3x3 filter applied with stride 2
7
7
A closer look at spatial dimensions:
![Page 27: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/27.jpg)
7x7 input (spatially) assume 3x3 filter applied with stride 2=> 3x3 output!
7
7
A closer look at spatial dimensions:
![Page 28: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/28.jpg)
7x7 input (spatially) assume 3x3 filter applied with stride 3?
7
7
A closer look at spatial dimensions:
![Page 29: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/29.jpg)
7x7 input (spatially) assume 3x3 filter applied with stride 3?
7
7
A closer look at spatial dimensions:
doesn’t fit!cannot apply 3x3 filter on 7x7 input with stride 3.
![Page 30: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/30.jpg)
In practice: Common to zero pad the border
e.g. input 7x73x3 filter, applied with stride 1pad with 1 pixel border => what is the output?
7x7 output!
0 0 0 0 0 0
0
0
0
0
![Page 31: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/31.jpg)
In practice: Common to zero pad the border
e.g. input 7x73x3 filter, applied with stride 1pad with 1 pixel border => what is the output?
7x7 output!in general, common to see CONV layers with stride 1, filters of size FxF, and zero-padding with (F-1)/2. (will preserve size spatially)e.g. F = 3 => zero pad with 1
F = 5 => zero pad with 2F = 7 => zero pad with 3
0 0 0 0 0 0
0
0
0
0
![Page 32: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/32.jpg)
Pooling layer- makes the representations smaller and more manageable- operates over each activation map independently:
![Page 33: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/33.jpg)
1 1 2 4
5 6 7 8
3 2 1 0
1 2 3 4
Single depth slice
x
y
max pool with 2x2 filters and stride 2 6 8
3 4
MAX POOLING
![Page 34: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/34.jpg)
Tensorflow implementation
• WeightInitialization
• ConvolutionandPooling• Convolutionlayer• Fullyconnectedlayer• ReadoutLayer
• Referenceandimagesource:https://www.tensorflow.org/get_started/mnist/pros
(Seesection‘BuildaMultilayerConvolutionalNetwork’)
![Page 35: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/35.jpg)
Input(placeholder)
tf.placeholder
xisplaceholderforinputimage.yislabel withone-hotrepresentation,soseconddimensionofyisequaltonumberofclasses.
None indicatesthatthefirstdimension,correspondingtothebatchsize,whichcanbeanysize.
![Page 36: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/36.jpg)
WeightInitialization
tf.truncated_normal
Thesevariablewillbeinitializedwhenuserrun‘tf.global_variables_initializer’.Nowtheyarejustnodesinagraphwithoutanyvalue.
![Page 37: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/37.jpg)
ConvolutionandPooling
Stridesis4-d,followingNHWCformat.(Num_samples xHeightxWidthxChannels)
Recallstridesandpadding.padding=‘SAME’meansapplypaddingtokeepoutputsizeassameasinputsize.
Conv2dpadswithzeros andmax_pool padswith–inf.tf.nn.conv2d
tf.nn.max_pool
![Page 38: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/38.jpg)
Convolutionlayer
tf.reshape
![Page 39: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/39.jpg)
Convolutionlayer
Seehowthecodecreatesamodelbywrappinglayers.Becareofshape ofeachlayer.-1meansmatchthesizeofthatdimensioniscomputedsothatthetotalsizeremainsconstant.
tf.reshape
![Page 40: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/40.jpg)
Reshape
tf.reshape
Forexample:
tensor‘t’is[[1,2],[3,4],[5,6],[7,8]],sothasshape[4,2]
(1) reshape(t,[2,4])è [[1,2,3,4],[5,6,7,8]]
(2) reshape(t,[-1,4])è [[1,2,3,4],[5,6,7,8]]
-1wouldbecomputedandbecomes‘2’
![Page 41: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/41.jpg)
Fullyconnectedlayer
Flattenallthe mapsandconnectthemwithfullyconnectedlayer.Again,becareofshape.
![Page 42: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/42.jpg)
ReadoutLayer
Usealayertomatchoutputsize.Done!
![Page 43: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/43.jpg)
TrainingandEvaluation(optional)
![Page 44: Tensorflow CNN turorial - 國立臺灣大學speech.ee.ntu.edu.tw/~tlkagk/courses/MLDS_2017/Lecture/... · 2017. 3. 25. · 2017/03/10. Lenet-5 [LeCunet al., 1998] ... ConvolutionLayer](https://reader034.fdocuments.in/reader034/viewer/2022051902/5ff23f8dd77ed856a9524efa/html5/thumbnails/44.jpg)
Recommendation
• Searchforeachfunction,andyou’llwhat’severythinggoingon.