Towards a Universal Gating Network for Mixtures of Experts

TL;DR

This paper introduces data-free methods for combining heterogeneous pre-trained neural networks using gating networks, achieving high accuracy and enabling applications in data-free regimes.

Contribution

It proposes novel data-free gating techniques for heterogeneous neural network combination, including a universal gating approach, advancing mixture of experts research.

Findings

Gating networks outperform other combination methods in accuracy.

Universal gating approach is effective for heterogeneous network merging.

Methods work well in data-free regimes.

Abstract

The combination and aggregation of knowledge from multiple neural networks can be commonly seen in the form of mixtures of experts. However, such combinations are usually done using networks trained on the same tasks, with little mention of the combination of heterogeneous pre-trained networks, especially in the data-free regime. This paper proposes multiple data-free methods for the combination of heterogeneous neural networks, ranging from the utilization of simple output logit statistics, to training specialized gating networks. The gating networks decide whether specific inputs belong to specific networks based on the nature of the expert activations generated. The experiments revealed that the gating networks, including the universal gating approach, constituted the most accurate approach, and therefore represent a pragmatic step towards applications with heterogeneous mixtures of experts in a data-free regime. The code for this project is hosted on github at https://github.com/cwkang1998/network-merging.