Rich Feature Construction for the Optimization-Generalization Dilemma

TL;DR

This paper introduces the Bonsai algorithm to construct rich feature representations that improve out-of-distribution generalization and optimization stability across multiple benchmarks.

Contribution

The paper presents a novel feature construction method, Bonsai, that enhances OoD generalization by providing rich initial representations and inductive biases.

Findings

Improves six OoD methods on ColoredMNIST benchmark

Outperforms existing methods on Camelyon17 dataset

Reduces variance and stabilizes hyperparameter tuning

Abstract

There often is a dilemma between ease of optimization and robust out-of-distribution (OoD) generalization. For instance, many OoD methods rely on penalty terms whose optimization is challenging. They are either too strong to optimize reliably or too weak to achieve their goals. We propose to initialize the networks with a rich representation containing a palette of potentially useful features, ready to be used by even simple models. On the one hand, a rich representation provides a good initialization for the optimizer. On the other hand, it also provides an inductive bias that helps OoD generalization. Such a representation is constructed with the Rich Feature Construction (RFC) algorithm, also called the Bonsai algorithm, which consists of a succession of training episodes. During discovery episodes, we craft a multi-objective optimization criterion and its associated datasets in a manner that prevents the network from using the features constructed in the previous iterations. During synthesis episodes, we use knowledge distillation to force the network to simultaneously represent all the previously discovered features. Initializing the networks with Bonsai representations consistently helps six OoD methods achieve top performance on ColoredMNIST benchmark. The same technique substantially outperforms comparable results on the Wilds Camelyon17 task, eliminates the high result variance that plagues other methods, and makes hyperparameter tuning and model selection more reliable.