There is a continuous reward and many discrete rewards. These rewards need to be merged into a new reward A. This reward A is designed as a weight to compute the weighted log-loss. In the past, I get the sum of the continuous reward and those discrete rewards. The trained results will tend to favor these discrete rewards.
1
-
$\begingroup$ Please clarify your specific problem or provide additional details to highlight exactly what you need. As it's currently written, it's hard to tell exactly what you're asking. $\endgroup$– Community BotCommented Aug 29 at 22:00
Add a comment
|