MACRO: A Meta-Algorithm for Conditional Risk Minimization

TL;DR

MACRO is a new meta-algorithm for conditional risk minimization that efficiently learns from sequential dependent data without storing all past data, providing theoretical guarantees and improved real-world prediction performance.

Contribution

It introduces MACRO, a practical meta-algorithm for CRM that overcomes previous theoretical limitations by maintaining and updating subroutines instead of storing all data.

Findings

MACRO achieves improved prediction accuracy on real data.

It provides learning guarantees for CRM in practical settings.

MACRO outperforms traditional non-conditional methods.

Abstract

We study conditional risk minimization (CRM), i.e. the problem of learning a hypothesis of minimal risk for prediction at the next step of sequentially arriving dependent data. Despite it being a fundamental problem, successful learning in the CRM sense has so far only been demonstrated using theoretical algorithms that cannot be used for real problems as they would require storing all incoming data. In this work, we introduce MACRO, a meta-algorithm for CRM that does not suffer from this shortcoming, but nevertheless offers learning guarantees. Instead of storing all data it maintains and iteratively updates a set of learning subroutines. With suitable approximations, MACRO applied to real data, yielding improved prediction performance compared to traditional non-conditional learning.