Model Agreement via Anchoring

TL;DR

This paper introduces a technique based on anchoring to bound model disagreement in various machine learning algorithms, showing how disagreement can be driven to zero by adjusting specific training parameters.

Contribution

The authors develop a general anchoring-based method to derive disagreement bounds and apply it to multiple algorithms, demonstrating how disagreement diminishes with key training parameters.

Findings

Disagreement bounds are established for stacked aggregation, gradient boosting, neural architecture search, and regression trees.

Disagreement can be driven to zero by increasing the number of models, iterations, architecture size, or tree depth.

Results generalize from one-dimensional to multi-dimensional regression with strongly convex loss.

Abstract

Numerous lines of aim to control $\textit{model disagreement}$ -- the extent to which two machine learning models disagree in their predictions. We adopt a simple and standard notion of model disagreement in real-valued prediction problems, namely the expected squared difference in predictions between two models trained on independent samples, without any coordination of the training processes. We would like to be able to drive disagreement to zero with some natural parameter(s) of the training procedure using analyses that can be applied to existing training methodologies. We develop a simple general technique for proving bounds on independent model disagreement based on $\textit{anchoring}$ to the average of two models within the analysis. We then apply this technique to prove disagreement bounds for four commonly used machine learning algorithms: (1) stacked aggregation over an arbitrary model class (where disagreement is driven to 0 with the number of models $k$ being stacked) (2) gradient boosting (where disagreement is driven to 0 with the number of iterations $k$) (3) neural network training with architecture search (where disagreement is driven to 0 with the size $n$ of the architecture being optimized over) and (4) regression tree training over all regression trees of fixed depth (where disagreement is driven to 0 with the depth $d$ of the tree architecture). For clarity, we work out our initial bounds in the setting of one-dimensional regression with squared error loss -- but then show that all of our results generalize to multi-dimensional regression with any strongly convex loss.