Statistics as a Way to Model Unknown Dynamics
I think distribution can be seen as a way to model unknown dynamics over infinitesimal amount of time.
[Read More]
Uninteneded Acceleration
Zeta Greedy Policy
An Alternative to Epsilon Greedy for Balancing Exploration and Exploitation
Exit
I want to write down the strategy an investigator used to manage his risk exposure when exploring the unknown. The question posed to him was how does he decide what to investigate?
[Read More]
Bootstrapping Vs. Perfection
One key idea in Reinforcement Learning is bootstrapping. The target might not be the real target for now, but it is getting incrementally closer to the real target. I’ll put this post here for now and elaborate later on.
[Read More]