3 ms·
It’s best for immediate rewards. If you have delayed rewards there is a paper on sampling from the “delay distribution” that solves this.
by orasis 2y 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.