dbn Package

dbn Package

harmonium Module

Code for exponential family harmoniums

Follows Welling, Rosen-Zvi, Hinton 2005

class infpy.dbn.harmonium.Harmonium(observed, hidden)[source]

Bases: object

An exponential family harmonium with hidden and observed variables

Follows Welling, Rosen-Zvi, Hinton 2005

lambda_hat(j, observed_values)[source]

The natural parameters of a hidden variable taking quadratic term for interactions into account

theta_hat(i, hidden_values)[source]

The natural parameters of an observed variable taking quadratic term for interactions into account

infpy.dbn.harmonium.contrastive_divergence_iteration(harmonium, observed_data, eta)[source]

One iteration of contrastive divergence

rbm Module

class infpy.dbn.rbm.RBM(H, V)[source]

Bases: object

A restricted boltzman machine

P_v(h)[source]

Compute P(v=1|h)

Q_h(v)[source]

Compute Q(h=1|v)

init_random(scale=1.0)[source]

Randomly initialise the biases and weights

update(v)[source]

Update a restricted boltzman machine

infpy.dbn.rbm.sigmoid(x)[source]

Table Of Contents

Previous topic

infpy Package

Next topic

decision Package

This Page