deep learning步骤
(deep=many hidden layers )
step1:定义一个function / neural network
1.参数:weight 和 bias
2.fully connect feedforward network
3.运算方法:矩阵运算(matrix operation)
4.分类:
常见deep learning的结构:AlexNet、VGG、GoogleNet、ResNet
step2:goodness of function
评价参数:total loss(通过 cross entropy|交叉熵 计算)、gradient descent(梯度下降)、backpropagation