【機器學習2021】6 Convolutional Neural Networks

卷積神經網路 (Convolutional Neural Networks, CNN)

Model Bias:

https://stackoverflow.com/questions/2480650/what-is-the-role-of-the-bias-in-neural-networks

I think that biases are almost always helpful. In effect, a bias value allows you to shift the activation function to the left or right, which may be critical for successful learning.

example of bias’ effect: https://stackoverflow.com/questions/38248657/why-is-a-simple-2-layer-neural-network-unable-to-learn-0-0-sequence#38253140

Since a neural network is performing multiplication between each layer, the weights have no impact, since any number times 0 is still 0.

Common Bias Set: 1 ----------> WHY bias is set 1 ??

Because by multiplying a bias by a weight, you can shift it by an arbitrary amount

其它解释： y = w X + b 中 w为weight，b为bias

Pooling – subsampling

convolution - pooling交替：

最主要的理由：减少运算量. 现在算力足够可以full conv

CNN Model

CNN 需要data augmentation: 训练时目标放大、旋转…