Realistic actor-critic: a framework for balance between value overestimation and underestimation

HIGHLIGHTS

• who: Sicen Li from the University of Leicester, United Kingdom have published the research work: Realistic Actor-Critic: A framework for balance between value overestimation and underestimation, in the Journal: (JOURNAL)
• what: The authors show empirically that RAC controls the standard deviation and the mean of value estimate bias to close to zero for most of the training. Empirically, the authors implement RAC with SAC (Haarnoja et_al, 2018) and TD3 (Fujimoto et_al, 2018) in continuous control benchmarks (OpenAI Gym Brockman et_al, 2016, MuJoCo Todorov et_al, 2012). This work propose to solve this problem by learning . . .
  

   

  

  If you want to have access to all the content you need to log in!

  Thanks :)

  Username or E-mail:

  If you don't have an account, you can create one here.