🏢 Gatsby Computational Neuroscience Unit, University College London
Near-Optimality of Contrastive Divergence Algorithms
·280 words·2 mins·
loading
·
loading
Machine Learning
Unsupervised Learning
🏢 Gatsby Computational Neuroscience Unit, University College London
Contrastive Divergence algorithms achieve near-optimal parameter estimation rates, matching the Cramér-Rao lower bound under specific conditions, as proven by a novel non-asymptotic analysis.