f7a1a77baf9102cbd20971c8edba71c4adad33d6,gpytorch/__init__.py,,monte_carlo_log_likelihood,#Any#Any#Any#Any#Any#,106

Before Change

                                                                variational_mean,
                                                                chol_var_covar)
    else:
        epsilon = Variable(train_covar.data.new(len(train_covar), num_trace_samples).normal_())
        samples = chol_var_covar.t().mm(epsilon)
        samples = samples + variational_mean.unsqueeze(1)
        log_likelihood = log_probability_func(samples, train_y)

After Change

    Args:
        - covar (matrix nxn) - Variable or LazyVariable representing the covariance matrix of the observations.
                               Usually, this is K + s*I, where s is the noise variance, and K is the prior covariance.
        - target (vector n) - Training label vector.

    Returns:
        - scalar - The marginal log likelihood of the data.

Instances