Why does randomizing samples of reinforcement learning model with a non-linear function approximator reduce variance?

While reading the DQN paper, I found that randomly selecting and learning samples reduced divergence in RL using a non-linear function approximator (e.g a neural network). So why does Reinforcement Learning using a non-linear function approximator diverge when using strongly correlated data as input?

Acoustic Input Data: Decibel or Pascals

In acoustics decibel levels were defined to solve an issue with showing values that are interpretive, understandable, and easy to communicate in contrast to intensity or pressure in Pascals. $dB = 10*\log({\frac{p^2}{p_{ref}^2}})$ This log scale helps human understanding of an acoustic signal because human hearing is capable of discerning the difference of about 1 dB…

Is there a reason to choose regular momentum over Nesterov momentum for neural networks?

I’ve been reading about Nesterov momentum from here and it seems like a nice improvement over regular momentum with no extra cost whatsoever. However, is this really the case? Are there instances where regular momentum performs better than Nesterov momentum or is Nesterov momentum performs at least as good as the regular momentum all the…

Predict a value calculated from a hash (and hash before that hash)

I have a hash, e.g. 552c51576e2a6dec462b0d56af8c31713e26a763c0e2a3d48ebbf4b99958eeba. This hash hashed again, is 39f20b3e5cc0181d6fbe2a4d52c7b3c315e8526e35da3cd0b18e25d10b670dc3. Now I only know the last hash 39f20b3e5cc0181d6fbe2a4d52c7b3c315e8526e35da3cd0b18e25d10b670dc3. It matches value 1.58. I do not know the first hash, as this is a security feature (that all hashs are generated and we go from top to bottom to not re-calculate the hashes). A…