CNN- Convolution Those are the network parameters to be learned 1 1000 0 Filter 1 010010 Matrix 001100 00010 Filter 2 010010 Matrix 00 010 6X 6 image Each filter detects a small Property 1 pattern (3 X 3) Sourceoftheslidehttp://219.216.82.193/cache/8/03/speech.ee.ntu.edu.tw/43149163c97eb6be7590e3d8de445a67/cnn.pdf
11 CNN – Convolution 1 0 0 0 0 1 0 1 0 0 1 0 0 0 1 1 0 0 1 0 0 0 1 0 0 1 0 0 1 0 0 0 1 0 1 0 6 x 6 image 1 -1 -1 -1 1 -1 -1 -1 1 Filter 1 -1 1 -1 -1 1 -1 -1 1 -1 Filter 2 … … Those are the network parameters to be learned. Matrix Matrix Each filter detects a small pattern (3 x 3). Property 1 Source of the slide: http://219.216.82.193/cache/8/03/speech.ee.ntu.edu.tw/43149163c97eb6be7590e3d8de445a67/CNN.pdf
cNN- Convolution Filter 1 stride= 1 10000 0 001 3 001 0 0001 01001 00000 00 01 6x 6 image 12 Sourceoftheslidehttp://219.216.82.193/cache/8/03/speech.ee.ntu.edu.tw/43149163c97eb6be7590e3d8de445a67/cnn.pdf
12 CNN – Convolution 1 0 0 0 0 1 0 1 0 0 1 0 0 0 1 1 0 0 1 0 0 0 1 0 0 1 0 0 1 0 0 0 1 0 1 0 6 x 6 image 1 -1 -1 -1 1 -1 -1 -1 1 Filter 1 3 -1 stride=1 Source of the slide: http://219.216.82.193/cache/8/03/speech.ee.ntu.edu.tw/43149163c97eb6be7590e3d8de445a67/CNN.pdf
cNN- Convolution Filter 1 If stride=2 10000|1 0 001 3 3 00110 0001 0100 00000 We set stride=1 below 00 01 6x 6 image 13 Sourceoftheslidehttp://219.216.82.193/cache/8/03/speech.ee.ntu.edu.tw/43149163c97eb6be7590e3d8de445a67/cnn.pdf
13 CNN – Convolution 1 0 0 0 0 1 0 1 0 0 1 0 0 0 1 1 0 0 1 0 0 0 1 0 0 1 0 0 1 0 0 0 1 0 1 0 6 x 6 image 1 -1 -1 -1 1 -1 -1 -1 1 Filter 1 3 -3 If stride=2 We set stride=1 below Source of the slide: http://219.216.82.193/cache/8/03/speech.ee.ntu.edu.tw/43149163c97eb6be7590e3d8de445a67/CNN.pdf
cNN- Convolution Filter 1 1|-1N stride= 1 M00001 0 00|1 3 3 00 10 0001 00000 3 0 3 0 00 00 01 3)(0 6x 6 image 3 2 Property 2 14 Sourceoftheslidehttp://219.216.82.193/cache/8/03/speech.ee.ntu.edu.tw/43149163c97eb6be7590e3d8de445a67/cnn.pdf
14 CNN – Convolution 1 0 0 0 0 1 0 1 0 0 1 0 0 0 1 1 0 0 1 0 0 0 1 0 0 1 0 0 1 0 0 0 1 0 1 0 6 x 6 image 1 -1 -1 -1 1 -1 -1 -1 1 Filter 1 3 -1 -3 -1 -3 1 0 -3 -3 -3 0 1 3 -2 -2 -1 stride=1 Property 2 Source of the slide: http://219.216.82.193/cache/8/03/speech.ee.ntu.edu.tw/43149163c97eb6be7590e3d8de445a67/CNN.pdf
cNN- Convolution 1 Filter 2 stride= 1 Do the same process for 100001 every filter 01|0010 001100 100010 010010 Feature 001010 Map 6x 6 image 10-43 4 X 4 image 15 Sourceoftheslidehttp://219.216.82.193/cache/8/03/speech.ee.ntu.edu.tw/43149163c97eb6be7590e3d8de445a67/cnn.pdf
15 CNN – Convolution 1 0 0 0 0 1 0 1 0 0 1 0 0 0 1 1 0 0 1 0 0 0 1 0 0 1 0 0 1 0 0 0 1 0 1 0 6 x 6 image 3 -1 -3 -1 -3 1 0 -3 -3 -3 0 1 3 -2 -2 -1 -1 1 -1 -1 1 -1 -1 1 -1 Filter 2 -1 -1 -1 -1 -1 -1 -2 1 -1 -1 -2 1 -1 0 -4 3 Do the same process for every filter stride=1 4 x 4 image Feature Map Source of the slide: http://219.216.82.193/cache/8/03/speech.ee.ntu.edu.tw/43149163c97eb6be7590e3d8de445a67/CNN.pdf