Python rank_update_mean_covariance_cholesky_lmbda_naive Examples

Example #1

File: test_adaptive_metropolis.py Project: swyoon/kernel_hmc

def test_rank_update_mean_covariance_cholesky_lmbda_cholupdate_equals_naive():
    if not low_rank_update_available:
        raise SkipTest("cholupdate not available")

    D = 3
    N = 100
    X = np.random.randn(N, D)

    mean = np.mean(X, 0)
    Sigma = np.cov(X.T)
    L = np.linalg.cholesky(Sigma)
    assert_allclose(np.dot(L, L.T), Sigma)

    # update with one more vector
    u = np.random.randn(D)
    lmbda = 0.1
    nu2 = 0.2
    gamma2 = 0.3

    m_test_naive, L_test_naive = rank_update_mean_covariance_cholesky_lmbda_naive(
        u, lmbda, mean, L, nu2=nu2, gamma2=gamma2)
    m_test, L_test = rank_update_mean_covariance_cholesky_lmbda_naive(
        u, lmbda, mean, L, nu2=nu2, gamma2=gamma2)

    assert_allclose(m_test_naive, m_test)
    assert_allclose(L_test_naive, L_test)

Example #2

File: test_adaptive_metropolis.py Project: swyoon/kernel_hmc

def test_rank_update_mean_covariance_cholesky_lmbda_naive():
    D = 3
    N = 100
    X = np.random.randn(N, D)

    mean = np.mean(X, 0)
    Sigma = np.cov(X.T)
    L = np.linalg.cholesky(Sigma)
    assert_allclose(np.dot(L, L.T), Sigma)

    # update with one more vector
    u = np.random.randn(D)
    lmbda = 0.1
    nu2 = 0.2
    gamma2 = 0.3

    updated_mean = (1 - lmbda) * mean + lmbda * u
    updated_Sigma = (1 - lmbda) * Sigma + lmbda * nu2 * np.outer(
        u - mean, u - mean) + np.eye(D) * gamma2
    updated_L = np.linalg.cholesky(updated_Sigma)

    m_test, L_test = rank_update_mean_covariance_cholesky_lmbda_naive(
        u, lmbda, mean, L, nu2=nu2, gamma2=gamma2)
    assert_allclose(updated_mean, m_test)
    assert_allclose(updated_L, L_test)

Example #3

File: test_adaptive_metropolis.py Project: littlefoxhome/kernel_hmc

def test_rank_update_mean_covariance_cholesky_lmbda_cholupdate_equals_naive():
    if not low_rank_update_available:
        raise SkipTest("cholupdate not available")
    
    D = 3
    N = 100
    X = np.random.randn(N, D)
    
    mean = np.mean(X, 0)
    Sigma = np.cov(X.T)
    L = np.linalg.cholesky(Sigma)
    assert_allclose(np.dot(L, L.T), Sigma)
    
    # update with one more vector
    u = np.random.randn(D)
    lmbda = 0.1
    nu2 = 0.2
    gamma2 = 0.3
    
    m_test_naive, L_test_naive = rank_update_mean_covariance_cholesky_lmbda_naive(u, lmbda, mean, L, nu2=nu2, gamma2=gamma2)
    m_test, L_test = rank_update_mean_covariance_cholesky_lmbda_naive(u, lmbda, mean, L, nu2=nu2, gamma2=gamma2)
    
    assert_allclose(m_test_naive, m_test)
    assert_allclose(L_test_naive, L_test)

Example #4

File: test_adaptive_metropolis.py Project: littlefoxhome/kernel_hmc

def test_rank_update_mean_covariance_cholesky_lmbda_naive():
    D = 3
    N = 100
    X = np.random.randn(N, D)
    
    mean = np.mean(X, 0)
    Sigma = np.cov(X.T)
    L = np.linalg.cholesky(Sigma)
    assert_allclose(np.dot(L, L.T), Sigma)
    
    # update with one more vector
    u = np.random.randn(D)
    lmbda = 0.1
    nu2 = 0.2
    gamma2 = 0.3
    
    updated_mean = (1 - lmbda) * mean + lmbda * u
    updated_Sigma = (1 - lmbda) * Sigma + lmbda * nu2 * np.outer(u - mean, u - mean) + np.eye(D) * gamma2
    updated_L = np.linalg.cholesky(updated_Sigma)
    
    m_test, L_test = rank_update_mean_covariance_cholesky_lmbda_naive(u, lmbda, mean, L, nu2=nu2, gamma2=gamma2)
    assert_allclose(updated_mean, m_test)
    assert_allclose(updated_L, L_test)