ELODI: Ensemble Logit Difference Inhibition for Positive-Congruent Training

TL;DR

ELODI is a novel training method that reduces negative flips in model updates by distilling ensemble logits into a single model, maintaining accuracy and lowering error rates without increasing inference costs.

Contribution

The paper introduces ELODI, a new ensemble distillation technique that effectively reduces negative flips while preserving accuracy, using a generalized logit difference inhibition objective.

Findings

ELODI achieves lower negative flip rates compared to existing methods.

The method maintains high accuracy during model updates.

ELODI reduces inference costs by using a single model after distillation.

Abstract

Negative flips are errors introduced in a classification system when a legacy model is updated. Existing methods to reduce the negative flip rate (NFR) either do so at the expense of overall accuracy by forcing a new model to imitate the old models, or use ensembles, which multiply inference cost prohibitively. We analyze the role of ensembles in reducing NFR and observe that they remove negative flips that are typically not close to the decision boundary, but often exhibit large deviations in the distance among their logits. Based on the observation, we present a method, called Ensemble Logit Difference Inhibition (ELODI), to train a classification system that achieves paragon performance in both error rate and NFR, at the inference cost of a single model. The method distills a homogeneous ensemble to a single student model which is used to update the classification system. ELODI also introduces a generalized distillation objective, Logit Difference Inhibition (LDI), which only penalizes the logit difference of a subset of classes with the highest logit values. On multiple image classification benchmarks, model updates with ELODI demonstrate superior accuracy retention and NFR reduction.