params['G_0'].set(0.10618535357378697) params['S'].set(0.10018501374500711) params['L'].set(np.array([0.92012063497620489, 0.079879365023795085])) params['T'].set(0)""" #run viterbi state_estimates, masteries = model.viterbi() #first val needs to be pulled off (I think) state_estimates = [x[1:] for x in state_estimates] time_estimates = get_mastery_time(state_estimates) trans_errs = [] for c in range(len(times)): trans_errs.append(time_estimates[c] - times[c]) model.load_test_split(X,P,False) trans_pred = model.get_predictions() trans_rmse = np.sqrt(np.mean( (trans_pred - X) ** 2)) #print trans_errs # Setup KT-IDEAL model pdictl = json.load(open("dump/PARAMS_simulated_trans_"+str(students)+"_second_ktideal_2states_1000iter.json","r")) model = KTIDEAL(X,P,0,0.15,False,False) params = model.get_parameters() #Set the learned model parameters for transition model for k, v in params.iteritems(): if k in pdictl[-1]: v.set(pdictl[-1][k]) else:
for c in range(loop): a = time.time() mcmc.MH(per_loop) b = time.time() print ("finished iteration: " + str((c + 1) * per_loop) + " in " + str(int(b - a)) + " seconds") end = time.time() print ("Finished burnin and " + str(num_iterations) + " iterations in " + str(int(end - start)) + " seconds.") folder = "plots_" + fname # plotting samples will also load the MAP estimates # mcmc.plot_samples(folder + "/", str(num_iterations) + '_iterations') # load up test data and run predictions model.load_test_split(Xtest, Ptest) pred = model.get_predictions() num = model.get_num_predictions() mast = model.get_mastery() err = pred - Xtest rmse = np.sqrt(np.sum(err ** 2) / num) errl = np.zeros(num) predl = np.zeros(num) mastl = np.zeros(num) xtestl = np.zeros(num) i = 0 for n in range(pred.shape[0]): for t in range(pred.shape[1]): if pred[n][t] == -1: