Add epsilon decay

[satwikkansal]

The current implementation could improve q-scores faster if we start from epsilon value of 0.95, and decay it gradually to 0.1.

Something like the following should do the trick

epsilon = min(0.1, epsilon * decay_factor)

[rushiv0609]

hi @satwikkansal , can i take this issue? This is an easy fix we will initialize the global epsilon parameter and update it in the update() function.

[satwikkansal]

Hi @rushiv0609,

Sure! I've assigned you to it. Please compare the results before and after the fix to make sure that it's an improvement and not a regression.

[rushiv0609]

#3