Brain-like approaches to unsupervised learning of hidden representations -- a comparative study

Unsupervised learning of hidden representations has been one of the most\nvibrant research directions in machine learning in recent years. In this work\nwe study the brain-like Bayesian Confidence Propagating Neural Network (BCPNN)\nmodel, recently extended to extract sparse distributed high-dimensional\nrepresentations. The usefulness and class-dependent separability of the hidden\nrepresentations when trained on MNIST and Fashion-MNIST datasets is studied\nusing an external linear classifier and compared with other unsupervised\nlearning methods that include restricted Boltzmann machines and autoencoders.\n