Open SamDuffield opened 2 months ago
Additionally considerations:
observation_noise_sd
(could be a different name) that corresponds to $\sigma$ in the predictive distribution $N( f \mid f^*, J^T \Sigma J + \sigma^2 \mathbb{I})$, details in https://arxiv.org/abs/1906.11537no_grad
to allow user more flexibility? Although majority of cases the function would still need to be wrapped in no_grad
to avoid memory usage from undesired gradients
Currently we have an API like
but perhaps and API like
might be cleaner as it provides a new function that retains the required signature of
f
. It's also better fitting with thetorch.func
API