# Rank consistent ordinal regression for neural networks with application to age estimation

## TL;DR

This paper introduces CORAL, a rank-consistent ordinal regression framework for neural networks that improves age estimation accuracy by ensuring monotonicity and consistency across classifiers, applicable to various architectures.

## Contribution

The paper proposes CORAL, a theoretically grounded ordinal regression method that guarantees rank-monotonicity and can be integrated with any neural network architecture.

## Key findings

- Significant reduction in age prediction error on face datasets.
- CORAL maintains rank consistency and confidence score reliability.
- Method is architecture-agnostic and improves existing ordinal regression models.

## Abstract

In many real-world prediction tasks, class labels include information about the relative ordering between labels, which is not captured by commonly-used loss functions such as multi-category cross-entropy. Recently, the deep learning community adopted ordinal regression frameworks to take such ordering information into account. Neural networks were equipped with ordinal regression capabilities by transforming ordinal targets into binary classification subtasks. However, this method suffers from inconsistencies among the different binary classifiers. To resolve these inconsistencies, we propose the COnsistent RAnk Logits (CORAL) framework with strong theoretical guarantees for rank-monotonicity and consistent confidence scores. Moreover, the proposed method is architecture-agnostic and can extend arbitrary state-of-the-art deep neural network classifiers for ordinal regression tasks. The empirical evaluation of the proposed rank-consistent method on a range of face-image datasets for age prediction shows a substantial reduction of the prediction error compared to the reference ordinal regression network.