Synthetic Data using LDA Generative Model

ldaSynthData.jl

using PyPlot;
using Distributions;

TOPIC_N = 5;
VOCABULARY_SIZE = 1000;
DOC_NUM = 100;
TERM_PER_DOC = 200;

X = zeros(DOC_NUM,VOCABULARY_SIZE);

phi=[];
for i=1:TOPIC_N
    push!(phi,rand(Dirichlet(VOCABULARY_SIZE,0.01)));
end

for i=1:DOC_NUM    
    theta=rand(Dirichlet(TOPIC_N,0.8));
    for j=1:TERM_PER_DOC
        z = rand(Multinomial(1,theta));
        z_assignment = 1;
        for k=1:TOPIC_N
            if(z[k]==1)
                break;
            end
            z_assignment+=1;
        end
        w = rand(Multinomial(1,phi[z_assignment]))
        w_assignment=1;
        for k=1:VOCABULARY_SIZE
            if(w[k]==1)
                break;
            end
            w_assignment+=1;
        end
        X[i,w_assignment]+=1;
    end
end

matshow(X)