I want to know, how we can estimate about best reward function (RF) from the list of RFs?
let suppose, I have two reward functions, which one can give the best performance.
1st RF
Loss = sum i=0 to N (Ai-Bi)
2nd RF
average loss
Loss = 1/N [sum i=0 to N (Ai-Bi)]
So from equations, how we can know about best RF?
Dear Mm Muhammad Muhammad,
I suggest you to see links and attached files on topic.
https://academic.oup.com/cercor/article/11/3/183/265535
https://deepblue.lib.umich.edu/bitstream/handle/2027.42/89705/jdsorg_1.pdf
https://medium.com/@BonsaiAI/deep-reinforcement-learning-models-tips-tricks-for-writing-reward-functions-a84fe525e8e0
https://towardsdatascience.com/the-complete-reinforcement-learning-dictionary-e16230b7d24e
https://blog.goodaudience.com/what-is-inverse-reinforcement-learning-e333228af146
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5359274/
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4763736/
Best regards
You can pick it up automatically by using Matlab code, see e.g., https://www.mathworks.com/matlabcentral/answers/109027-finding-maximum-value-and-it-s-location-from-the-matrix
https://www.mathworks.com/matlabcentral/answers/47428-to-find-the-maximum-value-in-a-matrix
- Badges
- Science topic