20c65469fe1e090f4d51322178655c7253b85e93,test/examples/simple_gp_regression_test.py,,test_posterior_latent_gp_and_likelihood_with_optimization,#,78

Before Change

    // We should learn optimal hyperparmaters
    // bias should be near 0
    params = dict(posterior_observation_model.named_parameter_groups())["params"]
    constant_value = params.constant_mean.data.squeeze()[0]
    assert(math.fabs(constant_value) < .05)

    // log_lengthscale should be near -1.4
    log_lengthscale_value = params.log_lengthscale.data.squeeze()[0]

After Change

    infer = Inference(prior_observation_model)
    posterior_observation_model = infer.run(train_x, train_y, optimize=True)
    test_function_predictions = posterior_observation_model(test_x)
    mean_abs_error = torch.mean(torch.abs(test_y - test_function_predictions.mean()))
    assert(mean_abs_error.data.squeeze()[0] < 0.01)

Instances