maxent_toolbox

Maximum Entropy toolbox for MATLAB

View the Project on GitHub orimaoz/maxent_toolbox

maxent.wangLandau

Description

Estimates the partition function and entropy of Boltzmann distributions on binary inputs using the Wang-Landau method.

Usage

model_out = maxent.wangLandau(model)
model_out = maxent.wangLandau(model,Name,Value,...)

Arguments

Mandatory arguments

  • model - Maximum entropy model as returned by the trainModel function.

Optional arguments (in the form of name,value pairs)

  • binsize - Bin size for the energy histogram (Default: 0.01).
  • depth - Accuracy parameter for the simulation (integer). Higher values mean a higher accuracy and longer runtime. The final accuracy is in the order of exp(2^-(depth-1)).
  • separation - Number of samples to skip for every sample obtained in the MCMC random walk. A larger value decorrelates the samples and provides more accurate results, but incurs a longer run-time.
  • savefile - will constantly save the state in this file, and try to resume from it if it already exists.
  • save_delay - delay between saves (in seconds).

Output

model_out - Model structure with two additional fields appended to it:

  • model_out.z - Log partition function of the model.
  • model_out.entropy - Entropy of the model (in bits).