No more meta-parameter tuning in unsupervised sparse feature learning

TL;DR

This paper introduces a meta-parameter free, fast unsupervised feature learning algorithm that optimizes sparsity in a novel way, achieving state-of-the-art results on STL-10 and producing discriminative, generalizable features.

Contribution

The paper presents a new unsupervised feature learning method that eliminates the need for meta-parameter tuning, simplifying the process and improving performance.

Findings

Achieves state-of-the-art performance on STL-10

Produces highly discriminative features

Generalizes well across datasets

Abstract

We propose a meta-parameter free, off-the-shelf, simple and fast unsupervised feature learning algorithm, which exploits a new way of optimizing for sparsity. Experiments on STL-10 show that the method presents state-of-the-art performance and provides discriminative features that generalize well.