I am looking some good material (lecture notes, slides or textbook) which explains the soft actor-critic method with the derivations. I will be happy for the suggestions.
$\begingroup$
$\endgroup$
1
-
$\begingroup$ Please, define "good" in "good material". Good is very subjective, unless explained. $\endgroup$– nbroCommented Sep 11 at 11:05
Add a comment
|
2 Answers
$\begingroup$
$\endgroup$
SAC is known for its stability and efficiency in continuous control tasks. You can directly read the original papers on SAC with detailed derivations and explanations which is Haarnoja et al (2018), Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor.
There's an extended version including additional details, practical implementation considerations, and applications in robotics. Haarnoja et al, (2019), Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor.
Lapan (2020), Deep Reinforcement Learning Hands-On. This book provides an accessible introduction to reinforcement learning and practical implementations, including SAC on page 618. It explains the method and its underlying theory, with code examples in PyTorch. This is useful if you prefer a practical, code-oriented approach.
$\begingroup$
$\endgroup$
2
SAC: Soft-Actor critic method is popular especially in the field of robotics as the author had worked on this method during his PhD while working in the field of robotics.
For SAC, you might find this blog helpful: SAC blog
Prior to SAC if you need to know the intuition behind AC then you can refer to this blog: AC blog
-
$\begingroup$ I have read this blog and almost half of the discussion is full of errors. $\endgroup$ Commented Sep 13 at 10:56
-
$\begingroup$ @DSPinfinity did you find errors in both the blogs i.e. SAC blog and AC blog? or was it only in SAC blog. $\endgroup$– avengerCommented Sep 15 at 11:48