Empirical Evaluation of Rectified Activations in Convolutional Network

TL;DR

This paper empirically evaluates various rectified activation functions in CNNs, showing that randomized leaky ReLUs improve performance and challenge the belief that sparsity is key for ReLU success.

Contribution

Introduces and assesses a new randomized leaky ReLU (RReLU), demonstrating its advantages over traditional activations on image classification tasks.

Findings

RReLU improves accuracy on CIFAR-100 without ensembles.

Non-zero negative slopes enhance CNN performance.

Deterministic negative slopes tend to overfit on small datasets.

Abstract

In this paper we investigate the performance of different types of rectified activation functions in convolutional neural network: standard rectified linear unit (ReLU), leaky rectified linear unit (Leaky ReLU), parametric rectified linear unit (PReLU) and a new randomized leaky rectified linear units (RReLU). We evaluate these activation function on standard image classification task. Our experiments suggest that incorporating a non-zero slope for negative part in rectified activation units could consistently improve the results. Thus our findings are negative on the common belief that sparsity is the key of good performance in ReLU. Moreover, on small scale dataset, using deterministic negative slope or learning it are both prone to overfitting. They are not as effective as using their randomized counterpart. By using RReLU, we achieved 75.68\% accuracy on CIFAR-100 test set without multiple test or ensemble.