Semiring Activation in Neural Networks

TL;DR

This paper introduces semiring-based nonlinear operators as trainable components in neural networks, expanding the algebraic framework and demonstrating their viability in various architectures through experimental validation.

Contribution

It presents a novel class of trainable nonlinear operators based on semirings, generalizing traditional activation functions and enabling new algebraic structures in neural network design.

Findings

Semiring operators can replace traditional activations in neural networks.

Semiring-based activations are viable in fully connected and convolutional neural networks.

Trade-offs and challenges of using semiring activations are discussed.

Abstract

We introduce a class of trainable nonlinear operators based on semirings that are suitable for use in neural networks. These operators generalize the traditional alternation of linear operators with activation functions in neural networks. Semirings are algebraic structures that describe a generalised notation of linearity, greatly expanding the range of trainable operators that can be included in neural networks. In fact, max- or min-pooling operations are convolutions in the tropical semiring with a fixed kernel. We perform experiments where we replace the activation functions for trainable semiring-based operators to show that these are viable operations to include in fully connected as well as convolutional neural networks (ConvNeXt). We discuss some of the challenges of replacing traditional activation functions with trainable semiring activations and the trade-offs of doing so.