Brain-like approaches to unsupervised learning of hidden representations -- a comparative study
This work addresses the need for comparative analysis of brain-inspired models in unsupervised learning, but it is incremental as it applies an existing method to standard datasets without major breakthroughs.
The paper tackled the problem of evaluating brain-like unsupervised learning models for extracting hidden representations by comparing the Bayesian Confidence Propagating Neural Network (BCPNN) to methods like restricted Boltzmann machines and autoencoders on MNIST and Fashion-MNIST datasets, finding that BCPNN produces sparse, high-dimensional representations with class-dependent separability.
Unsupervised learning of hidden representations has been one of the most vibrant research directions in machine learning in recent years. In this work we study the brain-like Bayesian Confidence Propagating Neural Network (BCPNN) model, recently extended to extract sparse distributed high-dimensional representations. The usefulness and class-dependent separability of the hidden representations when trained on MNIST and Fashion-MNIST datasets is studied using an external linear classifier and compared with other unsupervised learning methods that include restricted Boltzmann machines and autoencoders.