# Information-theoretical label embeddings for large-scale image classification

Source: [Information-theoretical label embeddings for large-scale image classification](<https://www.sota2.com/research/paper/68fa629c1e209131ce9fa2f5>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2025-10-25T20:22:26.903Z

Authors: Fran\\c{c}ois Chollet

Published: 2016-07-19T18:40:01.000Z

## Abstract

We present a method for training multi-label, massively multi-class image classification models, that is faster and more accurate than supervision via a sigmoid cross-entropy loss (logistic regression). Our method consists in embedding high-dimensional sparse labels onto a lower-dimensional dense sphere of unit-normed vectors, and treating the classification problem as a cosine proximity regression problem on this sphere. We test our method on a dataset of 300 million high-resolution images with 17,000 labels, where it yields considerably faster convergence, as well as a 7% higher mean average precision compared to logistic regression.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/1607.05691>)

[PDF](<https://arxiv.org/pdf/1607.05691.pdf>)