No more meta-parameter tuning in unsupervised sparse feature learning
This addresses the challenge of reducing manual tuning in feature learning for machine learning practitioners, though it appears incremental as it builds on existing sparsity optimization methods.
The paper tackled the problem of meta-parameter tuning in unsupervised sparse feature learning by proposing a meta-parameter free algorithm that optimizes for sparsity, achieving state-of-the-art performance on STL-10 with discriminative features that generalize well.
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.