cNN- Convolution 16
16 CNN – Convolution
CNN- Colorful image 1 Filter 1 Filter 2 Colorful image 10000 0100 0 fo 001100 10o010 010010 00 0 11 017 Sourceoftheslidehttp://219.216.82.193/cache/8/03/speech.ee.ntu.edu.tw/43149163c97eb6be d8de445a67/CNN. pdf
17 CNN – Colorful image 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 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 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 1 -1 -1 -1 1 -1 -1 -1 1 Filter 1 -1 1 -1 -1 1 -1 -1 1 -1 Filter 2 1 -1 -1 -1 1 -1 -1 -1 1 1 -1 -1 -1 1 -1 -1 -1 1 -1 1 -1 -1 1 -1 -1 1 -1 -1 1 -1 -1 1 -1 -1 1 -1 Colorful image Source of the slide: http://219.216.82.193/cache/8/03/speech.ee.ntu.edu.tw/43149163c97eb6be7590e3d8de445a67/CNN.pdf
Convolution V.S. Fully Conneted 00 團围 -1-1-1-1 0 00 00 000 0 0 1 0010 0 convolution Imade 0000 0010 Fully 00 connected 00010 0 0010 00 18 Sourceoftheslidehttp://219.216.82.193/cache/8/03/speech.ee.ntu.edu.tw/43149163 d8de445a67/CNN.pdf
18 Convolution v.s. Fully Conneted 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 image convolution -1 1 -1 -1 1 -1 -1 1 -1 1 -1 -1 -1 1 -1 -1 -1 1 1 x 2 x … … 36 x … … 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 Fullyconnected Source of the slide: http://219.216.82.193/cache/8/03/speech.ee.ntu.edu.tw/43149163c97eb6be7590e3d8de445a67/CNN.pdf
⑥MD-1 Filter 1 2: o 3 -1 4: 3 0000 3 010013口 00110 8: 2(-2-1 0001 010010 10: 001010 13 6x 6 image Less parameters 15 Only connect to 9 input, not fully 16: connected 19 Sourceoftheslidehttp://219.216.82.193/cache/8/03/speech.ee.ntu.edu.tw/43149163c97eb6be7590e3d8de445a67/cnn.pdf
19 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: 2: 3: … 7: 8: 9: … 13: 14: 15: … Only connect to 9 input, not fully connected 4: 10: 16: 1 0 0 0 0 1 0 0 0 0 1 1 3 Less parameters! Source of the slide: http://219.216.82.193/cache/8/03/speech.ee.ntu.edu.tw/43149163c97eb6be7590e3d8de445a67/CNN.pdf
⑥MD-1 Filter 1 2: 4 3 1|000c3 0 00 (3 00110 8: 0001 010010 001010 13 6 x 6 image Less parameters! 15 16: Shared weights Even less parameters 20 Sourceoftheslidehttp://219.216.82.193/cache/8/03/speech.ee.ntu.edu.tw/43149163c97eb6be7590e3d8de445a67/cnn.pdf
20 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 1 -1 -1 -1 1 -1 -1 -1 1 Filter 1 1 0 0 0 0 1 0 0 0 0 1 1 3 -1 Shared weights 6 x 6 image Less parameters! Even less parameters! 1: 2: 3: … 7: 8: 9: … 13: 14: 15: … 4: 16: Source of the slide: http://219.216.82.193/cache/8/03/speech.ee.ntu.edu.tw/43149163c97eb6be7590e3d8de445a67/CNN.pdf