def decompose(train_data, vad_data, weight = 20, num_factors = 100, num_iters = 50, lam = 1e-1, batch_size = 1000):
#model parameters
num_factors = num_factors
num_iters = num_iters
batch_size = batch_size
alpha = weight
n_jobs = 1
lam_theta = lam_beta = lam
print '********************** Factorizing using Matrix factorization **********************************'
S = content_wmf.linear_surplus_confidence_matrix(train_data, alpha=alpha)
U, V, vad_ndcg = content_wmf.factorize(S, num_factors, vad_data=vad_data, num_iters=num_iters,
init_std=0.01, lambda_U_reg=lam_theta, lambda_V_reg=lam_beta,
dtype='float32', random_state=98765, verbose=True,
recompute_factors=batched_inv_joblib.recompute_factors_batched,
batch_size=batch_size, n_jobs=n_jobs)
return U, V
#train the model
num_factors = 100
num_iters = 50
batch_size = 1000
n_jobs = 1
lam_theta = lam_beta = 1e-1 #grid search [1e-5, 1e-3, 1e-1, 1, 10]
best_ndcg = -np.inf
U_best = None
V_best = None
best_alpha = 0
#for alpha in [2, 5, 10, 20, 30, 50]:
for alpha in [20]:
S = content_wmf.linear_surplus_confidence_matrix(train_data,
alpha=alpha)
U, V, vad_ndcg = content_wmf.factorize(
S,
num_factors,
vad_data=vad_data,
num_iters=num_iters,
init_std=0.01,
lambda_U_reg=lam_theta,
lambda_V_reg=lam_beta,
dtype='float32',
random_state=98765,
verbose=True,
recompute_factors=batched_inv_joblib.recompute_factors_batched,
batch_size=batch_size,
n_jobs=n_jobs)