def test_grid():
model = grid_potts_model(4, 5, al_size=2, seed=0)
gt = model.infer(algorithm='path_dp')
result = model.infer(algorithm='mean_field', max_iter=1000)
assert_results_close(result, gt, log_pf_tol=5.0, mp_mse_tol=0.05)
def test_grid_4x4x2():
model = grid_potts_model(4, 4, al_size=2, seed=0)
max_lh_gt = model.max_likelihood(algorithm='path_dp')
max_lh_gt_value = np.log(model.evaluate(max_lh_gt))
max_lh_lower_bound = lp_relaxation(model).lower_bound
max_lh_upper_bound = lp_relaxation(model).upper_bound
assert (max_lh_lower_bound <= max_lh_gt_value <= max_lh_upper_bound)
def test_infer_compare_with_pairwise_grid_5x5():
pw_model = grid_potts_model(5, 5, al_size=2, seed=0)
true_pf = np.exp(pw_model.infer(algorithm='path_dp').log_pf)
nfg_model = inferlo.NormalFactorGraphModel.from_model(pw_model)
pf = infer_edge_elimination(nfg_model)
assert np.allclose(true_pf, pf)
def test_grid():
model = grid_potts_model(10, 5, al_size=2)
true_result = model.infer(algorithm='path_dp')
result_bp = BP.infer(model)
assert_results_close(true_result,
result_bp,
log_pf_tol=0.1,
mp_mse_tol=2e-3)
def test_cycle_relaxation_2x2_cycle():
"""
Sherali-Adams is exact on line graph.
"""
model = grid_potts_model(2, 2, al_size=2, seed=0)
max_lh_gt = model.max_likelihood(algorithm='path_dp')
sa_res = minimal_cycle(model)
max_lh_ub = sa_res.upper_bound
assert np.allclose(max_lh_ub, np.log(model.evaluate(max_lh_gt)))
def test_custom_dataset():
model = grid_potts_model(4, 3, al_size=4, seed=1)
true_result = model.infer(algorithm='path_dp')
model = GenericGraphModel.from_model(model)
dataset = Dataset(model=model,
true_log_pf=true_result.log_pf,
true_marginals=true_result.marg_prob,
name='test_dataset')
dataset_loader.save_custom_dataset(dataset)
reloaded_dataset = dataset_loader.load_custom_dataset('test_dataset')
assert reloaded_dataset.model.num_variables == model.num_variables
assert reloaded_dataset.model.factors[1].var_idx == model.factors[
1].var_idx
assert np.allclose(reloaded_dataset.model.factors[1].values,
model.factors[1].values)
assert np.allclose(reloaded_dataset.true_log_pf, dataset.true_log_pf)
assert np.allclose(reloaded_dataset.true_marginals, dataset.true_marginals)
assert reloaded_dataset.name == 'test_dataset'
def test_grid_approx():
model = grid_potts_model(5, 15, al_size=2, seed=0)
mp_true = model.infer(algorithm='path_dp').marg_prob
mp = model.infer(algorithm='message_passing').marg_prob
assert np.mean(np.square(mp_true - mp)) < 1e-4
def test_sample_grid2x20x2():
model = grid_potts_model(2, 20, al_size=2, seed=0)
true_marg_probs = model.infer(algorithm='path_dp').marg_prob
samples = model.sample(num_samples=1000)
check_samples(samples=samples, true_marg_probs=true_marg_probs, tol=5e-4)
def test_max_likelihood_grid_4x50x2():
model = grid_potts_model(4, 50, al_size=2, seed=0)
true_ml = model.max_likelihood(algorithm='path_dp')
ml = max_likelihood_junction_tree(model)
assert np.allclose(ml, true_ml)
def test_inference_grid_4x50x2():
model = grid_potts_model(4, 50, al_size=2, seed=0)
ground_truth = model.infer(algorithm='path_dp')
result = infer_junction_tree(model)
assert_results_close(result, ground_truth)
def test_grid_2x2x10():
model = grid_potts_model(2, 2, al_size=10, seed=0)
max_lh_gt = model.max_likelihood(algorithm='bruteforce')
max_lh = max_lh_path_dp(model)
assert np.allclose(max_lh, max_lh_gt)
def test_potts_grid():
model = grid_potts_model(6, 5, al_size=3)
true_result = model.infer(algo='parth_dp')
model = inferlo.GenericGraphModel.from_model(model)
bw_result = BackwardBucketElimination.infer(model)
assert_results_close(true_result, bw_result)
def test_grid_3x3():
model = grid_potts_model(3, 3, al_size=2, seed=123)
gt = model.infer(algorithm='bruteforce')
result = model.infer(algorithm='path_dp')
assert_results_close(result, gt)