Convolutional Neural Networks.pdf

所需积分/C币:13 2018-03-04 02:27:10 797KB PDF
17
收藏 收藏
举报

Convolutional Neural Networks are very similar to ordinary Neural Networks from the previous chapter: they are made up of neurons that have learnable weights and biases. Each neuron receives some inputs, performs a dot product and optionally follows it with a non-linearity.
● ● ● ● ● POO POOL POC RELU RELU RELU RELURELU RELU CONV CONV CONV CONV CONVCONV 32 ○○○ 32 0 0 axon from a neuron synapse 000 den drite cell body u;a:+b U11 2x;+b output axon activation u22 (W-F+2P)S+1 221121 01121-1111-30 01121-111-3 (F-1)/2 +1-(10-3+0/2+1-4.5W-F+2P)S+1=(10-3+0)/2+1=45 X. shape:(11,11,4) V ·VL0,0,O」-np.sum(ⅩL:5,;5, WO) V「1,0,0l-np.sum(X「2:7,:5,:1*W0)+b0 V[2,0,0]=np.sum(X[4:9,5,:]*W0)+b0 V[3,0,0]=mp.sum(X[6:11,;5,:]*W)+b0 0 WO. shape:(5, 5, 4) V[0,0.1]=np.sum(X:5,:5,:]*W1)+b1 V[1,O,1] (X[2:7,:5,:]*W) ·VL2,0,1」-mp.sum(X4:9,:5,;」*W1)+b1 V「3,0.11-np.sum(X「6:11,5,;1*W1)+b1 V[0,1,1]=np.sum(X[:5,2:7,:]*W1)+b1 ●V[2,3,1]=np.sum(X[:9,6:11,:]*W1)+b1 WI W2=W1-F+2P)S+1 H2=(H1-F+2Ps+1

...展开详情
试读 21P Convolutional Neural Networks.pdf
立即下载 低至0.43元/次 身份认证VIP会员低至7折
一个资源只可评论一次,评论内容不能少于5个字
上传资源赚积分or赚钱
    最新推荐
    Convolutional Neural Networks.pdf 13积分/C币 立即下载
    1/21
    Convolutional Neural Networks.pdf第1页
    Convolutional Neural Networks.pdf第2页
    Convolutional Neural Networks.pdf第3页
    Convolutional Neural Networks.pdf第4页
    Convolutional Neural Networks.pdf第5页

    试读结束, 可继续读2页

    13积分/C币 立即下载 >