wolffd@0: function [coef, C] = learn_AR(data, k)
wolffd@0: % Find the ML parameters of a vector autoregressive process of order k.
wolffd@0: % [coef, C] = learn_AR(k, data)
wolffd@0: % data{l}(:,t) = the observations at time t in sequence l
wolffd@0: 
wolffd@0: warning('learn_AR seems to be broken');
wolffd@0: 
wolffd@0: nex = length(data);
wolffd@0: obs = cell(1, nex);
wolffd@0: for l=1:nex
wolffd@0:   obs{l} = convert_to_lagged_form(data{l}, k);
wolffd@0: end
wolffd@0: 
wolffd@0: % The initial parameter values don't matter, since this is a perfectly observable problem.
wolffd@0: % However, the size of F must be set correctly.
wolffd@0: y = data{1};
wolffd@0: [s T] = size(y);
wolffd@0: coef = rand(s,s,k);
wolffd@0: C = rand_psd(s);
wolffd@0: [F,H,Q,R,initx,initV] = AR_to_SS(coef, C, y);
wolffd@0: 
wolffd@0: max_iter = 1;
wolffd@0: fully_observed = 1;
wolffd@0: diagQ = 0;
wolffd@0: diagR = 0;
wolffd@0: [F, H, Q, R, initx, initV, loglik] = ...
wolffd@0:     learn_kalman(obs, F, H, Q, R, initx, initV, max_iter, diagQ, diagR, fully_observed);
wolffd@0: 
wolffd@0: [coef, C] = SS_to_AR(F, Q, k);
wolffd@0: