Neuron 神经元
Activation Function 激活函数 4
Back Propagation
Gradient Descent 梯度下降
All data are feed to the altgorithm (Batch)
requires convex objective function
The local optimal is found with this method
pros and cons of gradient descent
Stochastic Gradient Descent 随机梯度下降
Every piece of data are feeded to the network and update the network
has higher fluctuation