Comment by orasis

2 years ago

It’s best for immediate rewards. If you have delayed rewards there is a paper on sampling from the “delay distribution” that solves this.

0 comments