Commits · 67435b574663a25b49263007e7e3b424cd7dd88c · Franck Dary / RL-Parsing

Jun 25, 2021
- States working for oracle learning · 67435b57
  Franck Dary authored 3 years ago
  
  67435b57
Jun 24, 2021
- Started to work on states · bbdc365a
  Franck Dary authored 3 years ago
  
  bbdc365a
Jun 17, 2021
- New parameters : Nb of unaplayable actions picked before breaking the analysis · ead830cc
  Maxime Petit authored 3 years ago
  
  ead830cc
Jun 11, 2021
- Scores are now printed in deterministic order · 85a275f0
  Franck Dary authored 3 years ago
  
  85a275f0
Jun 10, 2021
- Added suffix feature to neural network · cdfa91d8
  Franck Dary authored 3 years ago
  
  cdfa91d8
Jun 08, 2021
- Added TAG transition and tagparser transition set · fc820be0
  Franck Dary authored 3 years ago
  
  fc820be0
May 26, 2021
- Added arc swift transition system · 7cc3e7f1
  Franck Dary authored 3 years ago
  
  7cc3e7f1
May 12, 2021
- Added hyperparameters as program arguments · 2038e1bc
  Franck Dary authored 3 years ago
  
  2038e1bc
- Added multiple reward functions and added program argument to chose them · e8b9c9f0
  Franck Dary authored 3 years ago
  
  e8b9c9f0
May 06, 2021
- Added incremental mode (-i) : parser cannot see right context if it has not... · 62b38a8c
  Franck Dary authored 3 years ago
  
  Added incremental mode (-i) : parser cannot see right context if it has not been fixated (so only accessible after a BACK)
  62b38a8c
May 04, 2021
- Corrected path of train log file · fdfc07d3
  Franck Dary authored 3 years ago
  
  fdfc07d3
- Fixed missing argument · 157585cf
  Franck Dary authored 3 years ago
  
  157585cf
- Improved debug print · 22de30ac
  Franck Dary authored 3 years ago
  
  22de30ac
- modified reward for back · 394637aa
  Maxime Petit authored 3 years ago
  
  394637aa
Apr 28, 2021
- Added history of transitions in feature function · b699477b
  Franck Dary authored 3 years ago
  
  b699477b
Apr 27, 2021
- Added Back transitions, problem : system can cycle · bf43cbf5
  Franck Dary authored 3 years ago
  
  bf43cbf5
- Added history in config · 39e488e9
  Franck Dary authored 3 years ago
  
  39e488e9
- ProbaOracle and probaRandom get smaller with number of epochs · fb6f6ffa
  Franck Dary authored 3 years ago
  
  fb6f6ffa
Apr 26, 2021
- Added form in feature function, and added threshold for number of occurences in dicts · 7729907f
  Franck Dary authored 3 years ago
  
  7729907f
- Refactored feature extraction, and added extractFeature function to NeuralNetwork · 3d80804a
  Franck Dary authored 3 years ago
  
  3d80804a
Apr 24, 2021
- Added bootstrap mode for oracle training · b1c976ae
  Franck Dary authored 3 years ago
  
  b1c976ae
- Reduced probaRandom and probaORacle · e0926705
  Franck Dary authored 3 years ago
  
  e0926705
- In RL, consider non appliable action · 9547a7de
  Franck Dary authored 3 years ago
  
  9547a7de
- Added CUDA support and changed Neural network weight init to avoid Q value explosion in RL · 7ff12b50
  Franck Dary authored 3 years ago
  
  7ff12b50
- Improved Replay Memory · 18082daf
  Franck Dary authored 3 years ago
  
  18082daf
- Code refactoring and cosmetic changes to print · bc508cdb
  Franck Dary authored 3 years ago
  
  bc508cdb
Apr 20, 2021
- Working RL · a35103cd
  Franck Dary authored 3 years ago
  
  a35103cd
- Start RL · bb5b34c5
  Maxime Petit authored 3 years ago
  
  bb5b34c5
Apr 15, 2021
- Added way to decode model in main · bf44c7fe
  Franck Dary authored 3 years ago
  
  bf44c7fe
- Save best network during Train · 18803742
  Franck Dary authored 3 years ago
  
  18803742
- Cleaned main, put functions into Train.py and Decode.py · 67c67305
  Franck Dary authored 3 years ago
  
  67c67305
- Working neural network training with oracle · 74a8e16e
  Franck Dary authored 3 years ago
  
  74a8e16e