# Zero-Shot Feature Selection via Transferring Supervised Knowledge

## TL;DR

This paper introduces a zero-shot feature selection method that leverages class-semantic descriptions to transfer supervised knowledge to unseen concepts, improving feature selection when labeled data is scarce.

## Contribution

The authors propose a novel zero-shot feature selection approach using semantic descriptions and a center-characteristic loss to enhance discriminative feature learning.

## Key findings

- Effective in selecting features for unseen concepts
- Outperforms existing methods on real-world datasets
- Utilizes semantic descriptions for knowledge transfer

## Abstract

Feature selection, an effective technique for dimensionality reduction, plays an important role in many machine learning systems. Supervised knowledge can significantly improve the performance. However, faced with the rapid growth of newly emerging concepts, existing supervised methods might easily suffer from the scarcity and validity of labeled data for training. In this paper, the authors study the problem of zero-shot feature selection (i.e., building a feature selection model that generalizes well to "unseen" concepts with limited training data of "seen" concepts). Specifically, they adopt class-semantic descriptions (i.e., attributes) as supervision for feature selection, so as to utilize the supervised knowledge transferred from the seen concepts. For more reliable discriminative features, they further propose the center-characteristic loss which encourages the selected features to capture the central characteristics of seen concepts. Extensive experiments conducted on various real-world datasets demonstrate the effectiveness of the method.