Sida I. Wang, Percy Liang, Arun Chaganty.

####Estimating Mixture Models via Mixtures of Polynomials.

NIPS 2015.

Provide a list of polynomial moment constraints, and recovery parameters of mixture models. See our NIPS paper for details.

The ipython notebooks MixtureOfGaussians.ipynb and MixtureLinearRegressions.ipynb shows how to use polymom for these mixture models.

See the codalab worksheet for an executable version, and see mompy for our Generalized Moment Problem solver.

requires cvxopt and sympy.

mompy: is a package for building the moment matrix, solving the sdp (requires cvxopt), and extracting solutions

# construct the degree d moment matrix with the provided symbols
MM = mp.MomentMatrix(d, symbols, morder='grevlex') 
# generate an SDP and calls cvxopt
solsdp = mp.solvers.solve_basic_constraints(MM, constraints, 1e-8); 
sol = mp.extractors.extract_solutions_lasserre(MM, solsdp['x'], Kmax = k)

models: contains code to generate Gaussian, multiview, MLR models, sampling from them, and obtaining the list of variables.

archive: experimental stuff

algos: contains spectral methods for estimating mixture models