Learning RBM (Gibbs sampling)

17th-October-2020

At the limit of 𝑇𝑇 → ∞, 𝐯𝐯𝑇𝑇 is distributed 𝑝𝑝 (𝐯𝐯; θ Θ). Generate distribution 𝑝𝑝𝑇𝑇 (𝐯𝐯) close to - 𝑝𝑝 (𝐯𝐯; θ Θ), calculate approximate negative term.

Learning of RBM (Contrastive Divergence (Hinton, 2002))

What if I quit 𝑇𝑇 at finite times? It corresponds to the minimization of the objective function called Contrastive Divergence.