maxent_toolbox

Mandatory arguments

• samples - Set of samples to train the model on, in the format (ncells x nsamples).
• model - Maximum entropy model as returned by the createModel function. It does not matter whether or not this model has been trained because only the its moment-generating functions are used in order to compute the marginals
• probabilities - The values used to reweight each sample, this should be a vector of size (1 xnsamples).

Output

Example usage

% create an ME model
model = maxent.createModel(10,'ising');   

% train it on a set of samples
model = maxent.trainModel(model,samples);

% create a set of samples corresponding to all possible states of the system
npatterns = 2^model.ncells;
unique_words = logical(de2bi(0:(npatterns-1)))';
    
% get the probabilities of all the possible patterns
logprobs = maxent.getLogProbability(model,unique_words);
probabilities = exp(logprobs);

% taking the values of all the unique words, reweighted by their probabilities,
% is equivalent to taking the observables of the model E_p[h(x)]:
marginals = maxent.getWeightedMarginals(unique_words,model,probabilities);

% the code above is functionally equivalent to calling: 
% marginals = maxent.getMarginals(model)