Prune Responsibly

TL;DR

This paper highlights how neural network pruning can worsen performance disparities across classes, emphasizing the need for fairness-aware evaluation and decision-making in model compression.

Contribution

It introduces a framework incorporating fairness metrics into pruning decisions, addressing overlooked biases in accuracy-efficiency trade-offs.

Findings

Pruning can exacerbate class imbalance and bias.

A Pareto-based framework helps balance fairness and efficiency.

Many models show increased disparities after pruning.

Abstract

Irrespective of the specific definition of fairness in a machine learning application, pruning the underlying model affects it. We investigate and document the emergence and exacerbation of undesirable per-class performance imbalances, across tasks and architectures, for almost one million categories considered across over 100K image classification models that undergo a pruning process.We demonstrate the need for transparent reporting, inclusive of bias, fairness, and inclusion metrics, in real-life engineering decision-making around neural network pruning. In response to the calls for quantitative evaluation of AI models to be population-aware, we present neural network pruning as a tangible application domain where the ways in which accuracy-efficiency trade-offs disproportionately affect underrepresented or outlier groups have historically been overlooked. We provide a simple, Pareto-based framework to insert fairness considerations into value-based operating point selection processes, and to re-evaluate pruning technique choices.