Closed nabenabe0928 closed 1 year ago
Targets are a bunch of self-made benchmarks and NB201.
Stopping is based on Median pruning and promotion is based on ASHA. Apparently, stopping is better. After than, mostly 8 workers.
Stopping is based on Median pruning and promotion is based on ASHA. Apparently, stopping is better.
$n \in \{1,2,4,8,16\}$
Model-based Asynchronous Hyperparameter and Neural Architecture Search
ABOHB paper
Main points
Fantasizing
Fantasizing is to simply marginalize the unobserved objective values. Let a set of observations be:
and a set of pending observations be:
Furthermore, let a utility function for an acquisition function be $u(x|\cdot)$. Then the fantasizing is to compute the following marginalized acquisition function: