Optimizing for Generalization in Machine Learning

Introduction

Recent advances in the field of machine learning (ML) are revolutionizing an ever-growing variety of domains, ranging from statistical learning algorithms in computer vision and natural language processing all the way to reinforcement learning algorithms in autonomous driving and conversational AI.

The Generalization Mystery

However, many of these breakthroughs demonstrate phenomena that lack explanations, and sometimes even contradict conventional wisdom. Perhaps the greatest mystery of modern ML—and arguably, one of the greatest mysteries of all of modern computer science—is the question of generalization: why do these immensely complex prediction rules successfully apply to future unseen instances?

Importance of Understanding Generalization

Apart from the pure scientific curiosity it stimulates, I believe that this lack of understanding poses a significant obstacle to widening the applicability of ML to critical applications, like in healthcare or autonomous driving, where the cost of error is disastrous.

Project Goals

The broad goal of this project is to tackle the generalization mystery in the context of both statistical learning and reinforcement learning, focusing on optimization algorithms being the de facto contemporary standard in training learning models.

Methodology

Our methodology points out inherent shortcomings of widely accepted viewpoints with regard to the generalization of optimization-based learning algorithms. It takes a crucially different approach that targets the optimization algorithm itself.

Key Steps

1. Building bottom-up from fundamental and tractable optimization models.
2. Identifying intrinsic properties.
3. Developing algorithmic methodologies that enable optimization to effectively generalize in modern statistical- and reinforcement-learning scenarios.

Expected Outcomes

A successful outcome would not only lead to a timely and crucial shift in the way the research community approaches the generalization of contemporary optimization-based ML, but it may also significantly transform the way we develop practical, efficient, and reliable learning systems.