An empirical study of the relation between network architecture and complexity

TL;DR

This study empirically examines how different neural network architectures and capacities affect generalization performance as data complexity increases in an image classification task.

Contribution

It provides systematic insights into the relationship between network architecture, capacity, and data complexity, which was previously underexplored.

Findings

Network capacity influences generalization error as data complexity increases.

Different architectures respond variably to increasing class numbers.

Empirical evidence supports the importance of matching network capacity to data complexity.

Abstract

In this preregistration submission, we propose an empirical study of how networks handle changes in complexity of the data. We investigate the effect of network capacity on generalization performance in the face of increasing data complexity. For this, we measure the generalization error for an image classification task where the number of classes steadily increases. We compare a number of modern architectures at different scales in this setting. The methodology, setup, and hypotheses described in this proposal were evaluated by peer review before experiments were conducted.