Naseem Alsadi

Naseem Alsadi

Artificial Intelligence | Computer Architecture & Design | Blockchain

Deep Q

less than 1 minute read

Published:

  • Discount factor allows for rewards in the future to be slightly minimized.
  • The further you advance in time, the more divergence with rewards. (e.g coins might not be placed in the same place in the future).
  • We do not want machine to memorize a sequence, hence discounting rewards in the future will discourage it from pursuing and exact sequence.

Deep Q Learning Algorithm

My helpful screenshot

You May Also Enjoy

Neuro Fuzzy Logic Midterm Notes

less than 1 minute read

Published:

  • Supervised Learning algorithms require a target output
    • As opposed to unsupervised in which has no target output
    • Reinforcement is gives feedback
  • Underfitting can be a result of having a lack of hidden neurons or lack of data
  • Overtraining can be a result of too many hidden neurons
  • Each neuron has many inputs but only one output
  • Activation layer of hidden layer must be nonlinear
    • Other layers (I/O) can be linear

Cache

3 minute read

Published:

Memory Types

Static RAM