SAU: Smooth activation function using convolution with approximate identities
This provides a smooth alternative for activation functions in neural networks, but it is incremental as it builds on existing smoothing techniques.
The paper tackled the problem of non-differentiable activation functions like ReLU by proposing Smooth Activation Unit (SAU), a smooth approximation using convolution with approximate identities, resulting in a 5.12% improvement with ShuffleNet V2 on CIFAR100.
Well-known activation functions like ReLU or Leaky ReLU are non-differentiable at the origin. Over the years, many smooth approximations of ReLU have been proposed using various smoothing techniques. We propose new smooth approximations of a non-differentiable activation function by convolving it with approximate identities. In particular, we present smooth approximations of Leaky ReLU and show that they outperform several well-known activation functions in various datasets and models. We call this function Smooth Activation Unit (SAU). Replacing ReLU by SAU, we get 5.12% improvement with ShuffleNet V2 (2.0x) model on CIFAR100 dataset.