Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on Mar 10, 2018

• Updated OpenAI Gym dependency in README

  andreimuntean

  committed
  50a71b0

  Copy full SHA for 50a71b0

Commits on Mar 26, 2017

• Updated dependencies in README

  andreimuntean

  committed
  59ecdb5

  Copy full SHA for 59ecdb5

• Wrote a setup script

  andreimuntean

  committed
  b7c3f71

  Copy full SHA for b7c3f71

Commits on Mar 11, 2017

• Elaborated the preprocessing section in the README

  andreimuntean

  authored
  ed2cbc7

  Copy full SHA for ed2cbc7

• Updated the README

  andreimuntean

  authored
  1d4e1bd

  Copy full SHA for 1d4e1bd

Commits on Mar 9, 2017

• Added support for OpenAI Gym 0.8

  andreimuntean

  committed
  8ce43b4

  Copy full SHA for 8ce43b4

Commits on Feb 23, 2017

• Corrected flag description

  andreimuntean

  committed
  f5e5b6f

  Copy full SHA for f5e5b6f

• Added support for TensorFlow 1.0 and implemented reward clipping

  andreimuntean

  committed
  6b975f5

  Copy full SHA for 6b975f5

Commits on Feb 14, 2017

• No longer set reward to -1 when losing a life while training

  andreimuntean

  committed
  9500bdf

  Copy full SHA for 9500bdf

• While training, loss of life is now treated as end of episode

  andreimuntean

  committed
  0646f43

  Copy full SHA for 0646f43

• No longer cropping observations. Will take longer for agents to start learning but will still learn optimal policies

  andreimuntean

  committed
  8c57cb0

  Copy full SHA for 8c57cb0

Commits on Feb 13, 2017

• Changed the activation function from ELU to ReLU

  andreimuntean

  committed
  ec7460e

  Copy full SHA for ec7460e

Commits on Feb 12, 2017

• Images are now cropped and rescaled to 47x47 & simplified color removal

  andreimuntean

  committed
  803b222

  Copy full SHA for 803b222

Commits on Feb 11, 2017

• Shell script now correctly sets session name

  andreimuntean

  committed
  0ad72ef

  Copy full SHA for 0ad72ef

• Fixed policy loss, clarified (and corrected) description of _apply_discount & added total loss to summaries

  andreimuntean

  committed
  469582b

  Copy full SHA for 469582b

• Fixed entropy regularization

  andreimuntean

  committed
  ecbafc9

  Copy full SHA for ecbafc9

Commits on Feb 10, 2017

• Set bias terms to 0

  andreimuntean

  committed
  5379d0f

  Copy full SHA for 5379d0f

• Added summaries for the environment & GPU no longer explicitly hidden since it does not always work

  andreimuntean

  committed
  651b9b7

  Copy full SHA for 651b9b7

• GPUs are now explicitly disabled

  andreimuntean

  committed
  8c696fd

  Copy full SHA for 8c696fd

• Finished implementing the model

  andreimuntean

  committed
  d4e9854

  Copy full SHA for d4e9854

Commits on Feb 9, 2017

• Observations are now resized with SciPy & removed observations_per_state flag since the RNN already has short-term memory

  andreimuntean

  committed
  5655b7f

  Copy full SHA for 5655b7f

• Improved code clarity & added a method for estimating state values

  andreimuntean

  committed
  8ac9e90

  Copy full SHA for 8ac9e90

• Implemented some methods for the policy network

  andreimuntean

  committed
  5fc5988

  Copy full SHA for 5fc5988

Commits on Feb 8, 2017

• Created the policy network, switched max_gradient flag with max_gradient_norm & changed frame interpolation code

  andreimuntean

  committed
  30b4244

  Copy full SHA for 30b4244

Commits on Feb 7, 2017

• Added a flag for entropy regularization & started implementing the agent

  andreimuntean

  committed
  ce75ebe

  Copy full SHA for ce75ebe

• Implemented worker threads

  andreimuntean

  committed
  1c2919b

  Copy full SHA for 1c2919b

Commits on Feb 6, 2017

• Configured thread cluster & implemented the environment wrapper

  andreimuntean

  committed
  7306869

  Copy full SHA for 7306869

Commits on Feb 5, 2017

• Added a flag for displaying the game screen of each agent

  andreimuntean

  committed
  3377d68

  Copy full SHA for 3377d68

• Added missing import

  andreimuntean

  committed
  ef0fac6

  Copy full SHA for ef0fac6

• Updated shell script and started programming

  andreimuntean

  committed
  6e2710c

  Copy full SHA for 6e2710c

• Created a shell script for initializing the tmux session

  andreimuntean

  committed
  14f3d6a

  Copy full SHA for 14f3d6a

• Initial commit

  andreimuntean

  committed
  aa06f4c

  Copy full SHA for aa06f4c