Expanded Gating Ranges Improve Activation Functions

TL;DR

This paper proposes expanding the gating range of activation functions using arctan, demonstrating improved performance of various self-gated activations and Gated Linear Units in transformer models.

Contribution

It introduces a trainable parameter to expand gating ranges beyond zero and one, enhancing the effectiveness of self-gated activation functions.

Findings

Expanded gating improves activation function performance.

Empirical results show outperforming existing functions in transformers.

Expanded gating benefits Gated Linear Units.

Abstract

Activation functions are core components of all deep learning architectures. Currently, the most popular activation functions are smooth ReLU variants like GELU and SiLU. These are self-gated activation functions where the range of the gating function is between zero and one. In this paper, we explore the viability of using arctan as a gating mechanism. A self-gated activation function that uses arctan as its gating function has a monotonically increasing first derivative. To make this activation function competitive, it is necessary to introduce a trainable parameter for every MLP block to expand the range of the gating function beyond zero and one. We find that this technique also improves existing self-gated activation functions. We conduct an empirical evaluation of Expanded ArcTan Linear Unit (xATLU), Expanded GELU (xGELU), and Expanded SiLU (xSiLU) and show that they outperform existing activation functions within a transformer architecture. Additionally, expanded gating ranges show promising results in improving first-order Gated Linear Units (GLU).