test_rbm.py 文件源码

def test_gibbs_smoke():
    # Check if we don't get NaNs sampling the full digits dataset.
    # Also check that sampling again will yield different results.
    X = Xdigits
    rbm1 = BernoulliRBM(n_components=42, batch_size=40,
                        n_iter=20, random_state=42)
    rbm1.fit(X)
    X_sampled = rbm1.gibbs(X)
    assert_all_finite(X_sampled)
    X_sampled2 = rbm1.gibbs(X)
    assert_true(np.all((X_sampled != X_sampled2).max(axis=1)))