WeightScale: Interpreting Weight Change in Neural Networks

TL;DR

This paper introduces WeightScale, a method to interpret neural network learning by analyzing layer-wise weight changes, revealing insights into how networks learn across different depths and task complexities.

Contribution

The paper proposes a novel approach combining dimensionality reduction and clustering to interpret weight dynamics in very deep neural networks.

Findings

Layer-wise weight change correlates with task complexity.

Deeper layers exhibit distinct learning patterns based on task difficulty.

The method scales effectively to deep networks.

Abstract

Interpreting the learning dynamics of neural networks can provide useful insights into how networks learn and the development of better training and design approaches. We present an approach to interpret learning in neural networks by measuring relative weight change on a per layer basis and dynamically aggregating emerging trends through combination of dimensionality reduction and clustering which allows us to scale to very deep networks. We use this approach to investigate learning in the context of vision tasks across a variety of state-of-the-art networks and provide insights into the learning behavior of these networks, including how task complexity affects layer-wise learning in deeper layers of networks.