Activation Function
Activation Functions
Activation function or transfer function is used to determine the output of a neuron or node. It is a mathematical gate in between the input feeding the current neuron and its output going to the next layer.
In deep learning, we commonly use non-linear Activation Function include
Sigmoid: Output limit to [0 1]. But gives gradient vanishing problem, not used anymore
ReLU(rectified linear unit): most commonly used in CNN (hidden layers)
Others: Tanh, Leaky ReLU, Maxout...
Output Activation Function
These functions are transformations we apply to vectors coming out from CNNs ( s ) before the loss computation. [ reference ]
Sigmoid
It squashes a vector in the range (0, 1). It is applied independently to each element of ss sisi. It’s also called logistic function.
Softmax
Where sjsj are the scores inferred by the net for each class in CC. Note that the Softmax activation for a class sisi depends on all the scores in ss.
Last updated