Recurrent Neural Networks with Flexible Gates using Kernel Activation Functions

TL;DR

This paper introduces a flexible gating mechanism for recurrent neural networks using kernel activation functions, improving accuracy and training efficiency on sequential data tasks.

Contribution

It proposes a novel gating architecture with kernel activation functions and residual connections, enhancing model flexibility and performance.

Findings

Improved accuracy on sequential MNIST variants.

Faster training convergence with the new gating method.

Negligible increase in computational cost.

Abstract

Gated recurrent neural networks have achieved remarkable results in the analysis of sequential data. Inside these networks, gates are used to control the flow of information, allowing to model even very long-term dependencies in the data. In this paper, we investigate whether the original gate equation (a linear projection followed by an element-wise sigmoid) can be improved. In particular, we design a more flexible architecture, with a small number of adaptable parameters, which is able to model a wider range of gating functions than the classical one. To this end, we replace the sigmoid function in the standard gate with a non-parametric formulation extending the recently proposed kernel activation function (KAF), with the addition of a residual skip-connection. A set of experiments on sequential variants of the MNIST dataset shows that the adoption of this novel gate allows to improve accuracy with a negligible cost in terms of computational power and with a large speed-up in the number of training iterations.