Learning (fine adjustment) of DBN

(21th-October-2020)

1. Adjustment to maximize the likelihood of the entire DBN 2. Copy it to another deterministic structure (Neural Net etc.), and perform supervised learning in it.

Example of generation model

Deep Boltzmann Machine (MNIST) • Learn two- and three-tier DBM models with MNIST (Handwritten Numerical Data Set) • Generate 𝑝𝑝 (𝐯𝐯) samples using Gibbs sampling.