Trading off Consistency and Dimensionality of Convex Surrogates for Multiclass Classification

Authors: Enrique Nueve, Dhamma Kimpara, Bo Waggoner, Jessica Finocchiaro

NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical We investigate two approaches for trading off consistency and dimensionality in multiclass classification while using a convex surrogate loss. We first formalize partial consistency when the optimized surrogate has dimension d n. We then check if partial consistency holds under a given embedding and low-noise assumption, providing insight into when to use a particular embedding into Rd. Finally, we present a new method to construct (fully) consistent losses with d n out of multiple problem instances. This paper does not include experiments requiring code.
Researcher Affiliation Academia Enrique Nueve Department of Computer Science University of Colorado Boulder enrique.nueveiv@colorado.edu Bo Waggoner Department of Computer Science University of Colorado Boulder bwag@colorado.edu Dhamma Kimpara Department of Computer Science University of Colorado Boulder dhamma.kimpara@colorado.edu Jessie Finocchiaro Department of Computer Science Boston College finocch@bc.edu
Pseudocode Yes Algorithm 1 Elicit mode via comparisons and the d-Cross Polytopes
Open Source Code No This paper does not include experiments requiring code.
Open Datasets No The paper is primarily theoretical and does not involve empirical experiments that would require training on datasets.
Dataset Splits No The paper is primarily theoretical and does not involve empirical experiments, therefore no dataset splits for validation are described.
Hardware Specification No The paper does not include experiments.
Software Dependencies No The paper does not include experiments.
Experiment Setup No The paper does not include experiments.