1
$\begingroup$

I am not able to understand why expected sarsa algorithm is an off-policy learning method? How is it that behaviour and target policies can be different? Also please tell under what situation and conditions is expected sarsa equivalent to Q-learning?

$\endgroup$