Activations Through Extensions: A Framework To Boost Performance Of Neural Networks

TL;DR

This paper introduces a unifying framework for activation functions in neural networks, proposing novel extension techniques that improve performance with minimal additional complexity, validated through theoretical analysis and empirical experiments.

Contribution

It presents a new framework that unifies existing activation functions and introduces extensions that enhance neural network performance with low overhead.

Findings

Extensions improve neural network accuracy

Performance gains observed in time-series tasks

Minimal additional computational cost

Abstract

Activation functions are non-linearities in neural networks that allow them to learn complex mapping between inputs and outputs. Typical choices for activation functions are ReLU, Tanh, Sigmoid etc., where the choice generally depends on the application domain. In this work, we propose a framework/strategy that unifies several works on activation functions and theoretically explains the performance benefits of these works. We also propose novel techniques that originate from the framework and allow us to obtain ``extensions'' (i.e. special generalizations of a given neural network) of neural networks through operations on activation functions. We theoretically and empirically show that ``extensions'' of neural networks have performance benefits compared to vanilla neural networks with insignificant space and time complexity costs on standard test functions. We also show the benefits of neural network ``extensions'' in the time-series domain on real-world datasets.