Norm-preserving Orthogonal Permutation Linear Unit Activation Functions (OPLU)
This is an incremental improvement for deep learning practitioners, potentially aiding in training deep and recurrent networks.
The authors tackled the problem of training deep and recurrent neural networks by proposing the OPLU activation function, which ensures norm preservation of backpropagated gradients, and it showed similar performance to tanh and ReLU on two toy problems.
We propose a novel activation function that implements piece-wise orthogonal non-linear mappings based on permutations. It is straightforward to implement, and very computationally efficient, also it has little memory requirements. We tested it on two toy problems for feedforward and recurrent networks, it shows similar performance to tanh and ReLU. OPLU activation function ensures norm preservance of the backpropagated gradients, therefore it is potentially good for the training of deep, extra deep, and recurrent neural networks.