Reformulating van Rijsbergen's $F_{\beta}$ metric for weighted binary cross-entropy

TL;DR

This paper introduces a novel approach to incorporate van Rijsbergen's $F_{eta}$ metric into binary cross-entropy loss by reformulating it for dynamic weighting, improving classification performance especially on imbalanced data.

Contribution

It reformulates the $F_{eta}$ metric for distributional assumptions and integrates it into a weighted binary cross-entropy, enabling better model performance and interpretability.

Findings

14% boost in $F_1$ score on IMDB data

Improved results over baseline on balanced and imbalanced classes

Enhanced interpretability of model performance

Abstract

The separation of performance metrics from gradient based loss functions may not always give optimal results and may miss vital aggregate information. This paper investigates incorporating a performance metric alongside differentiable loss functions to inform training outcomes. The goal is to guide model performance and interpretation by assuming statistical distributions on this performance metric for dynamic weighting. The focus is on van Rijsbergens $F_{\beta}$ metric -- a popular choice for gauging classification performance. Through distributional assumptions on the $F_{\beta}$, an intermediary link can be established to the standard binary cross-entropy via dynamic penalty weights. First, the $F_{\beta}$ metric is reformulated to facilitate assuming statistical distributions with accompanying proofs for the cumulative density function. These probabilities are used within a knee curve algorithm to find an optimal $\beta$ or $\beta_{opt}$. This $\beta_{opt}$ is used as a weight or penalty in the proposed weighted binary cross-entropy. Experimentation on publicly available data along with benchmark analysis mostly yields better and interpretable results as compared to the baseline for both imbalanced and balanced classes. For example, for the IMDB text data with known labeling errors, a 14% boost in $F_1$ score is shown. The results also reveal commonalities between the penalty model families derived in this paper and the suitability of recall-centric or precision-centric parameters used in the optimization. The flexibility of this methodology can enhance interpretation.