Commits · aa5b38f1cb9bbb996b58826fff32658a2f9a39b4 · Nikolai.Hartmann / KerasROOTClassification

Aug 21, 2018
- using new transform function in plotting · aa5b38f1
  Nikolai.Hartmann authored 6 years ago
  
  aa5b38f1
- introducing transform member function · 1ad29ff9
  Nikolai.Hartmann authored 6 years ago
  
  1ad29ff9
Aug 20, 2018
- not working yet · 5a494d99
  Nikolai.Hartmann authored 6 years ago
  
  5a494d99
- apply total weight where nescessary · c6e649fc
  Nikolai.Hartmann authored 6 years ago
  
  c6e649fc
- fixing skip_activation mode · cf8f14ef
  Nikolai.Hartmann authored 6 years ago
  
  cf8f14ef
- restructure plot_score function · 1fce12f7
  Nikolai authored 6 years ago
  
  1fce12f7
Aug 17, 2018
- turn back on division by mean weight · cb83fad4
  Nikolai.Hartmann authored 6 years ago
  
  cb83fad4
- make total weight more consistent · 0c201c50
  Nikolai authored 6 years ago
  
  0c201c50
- option to evaluate before activation · 391e7e92
  Nikolai authored 6 years ago
  
  391e7e92
Aug 16, 2018
- python3 compatibility · 290a8013
  Nikolai authored 6 years ago
  
  290a8013
- model plotting and alternative significance plot · ae824de0
  Nikolai.Hartmann authored 6 years ago
  
  ae824de0
- Improve loading from dir · bf50acf0
  Nikolai.Hartmann authored 6 years ago
  
  * Call _init_from_args instead of init in subclasses * Store project_type variable in ClassificationProjectRNN * Introduce load_from_dir function to automatically choose correct class
  bf50acf0
- normalize weights for default training as well · 4ac1c6b5
  Nikolai authored 6 years ago
  
  4ac1c6b5
- try normalising weights · 6a2f9292
  Nikolai authored 6 years ago
  
  6a2f9292
- put model checkpoint/reload weights into separate function · 46cb66fa
  Nikolai authored 6 years ago
  
  46cb66fa
Aug 15, 2018
- Merge branch 'dev-rnn' · f14c65cd
  Nikolai authored 6 years ago
  
  f14c65cd
- (hopefully) correct treatment of class_weight · c3855055
  Nikolai authored 6 years ago
  
  c3855055
- little fix · d92d5fbc
  Nikolai authored 6 years ago
  
  d92d5fbc
Aug 14, 2018
- load and plot before training RNN · beee5503
  Nikolai.Hartmann authored 6 years ago
  
  beee5503
- fix write_friend_tree · 4c64d916
  Nikolai.Hartmann authored 6 years ago
  
  4c64d916
- evaluate RNN working · 7c1241ff
  Nikolai.Hartmann authored 6 years ago
  
  7c1241ff
- model and training for RNN working · d09ab3d6
  Nikolai.Hartmann authored 6 years ago
  
  d09ab3d6
- data structuring for RNN working · fe84f351
  Nikolai.Hartmann authored 6 years ago
  
  fe84f351
- Merge branch 'master' into dev-rnn · 3d8acb8b
  Nikolai.Hartmann authored 6 years ago
  
  3d8acb8b
- remove validate and plot all from train · aa3ffe0a
  Nikolai.Hartmann authored 6 years ago
  
  aa3ffe0a
- put score evaluation in separate function · f8a5e443
  Nikolai.Hartmann authored 6 years ago
  
  f8a5e443
- correct quantile range for input plots · b9419f4d
  Nikolai.Hartmann authored 6 years ago
  
  b9419f4d
- making balanced training more efficient · 5b541afd
  Nikolai authored 6 years ago
  
  5b541afd
- starting to develop yield_batch function for RNN · f52e8ad9
  Nikolai authored 6 years ago
  
  f52e8ad9
- starting rnn wrapper · 9994f0ad
  Nikolai authored 6 years ago
  
  9994f0ad
- making balanced training more efficient · 6409522c
  Nikolai authored 6 years ago
  
  6409522c
Aug 13, 2018
- Merge remote-tracking branch 'origin/master' · 71b89893
  Nikolai.Hartmann authored 6 years ago
  
  71b89893
- plot train and test for ROC by default · e543ec8d
  Nikolai authored 6 years ago
  
  e543ec8d
- commenting number of cores for session · cb732b1c
  Nikolai authored 6 years ago
  
  cb732b1c
Aug 12, 2018
- staring eval_model script · 1f5ff1b3
  Nikolai authored 6 years ago
  
  1f5ff1b3
- allow both None and 0 for deactivating dropout · f3ce92d9
  Nikolai authored 6 years ago
  
  f3ce92d9
Aug 10, 2018
- try to avoid copying also when fitting scaler · ae08f994
  Nikolai.Hartmann authored 6 years ago
  
  ae08f994
- further memory improvements · 99b7dc94
  Nikolai.Hartmann authored 6 years ago
  
  99b7dc94
- Memory improvements · 0e1202f5
  Nikolai.Hartmann authored 6 years ago
  
  * stop option for reading a maximum number of events for train/test * inplace scaling
  0e1202f5
- Experimental support for initialising from pandas DataFrame (memory intense ...) · beea1cfe
  Nikolai.Hartmann authored 6 years ago
  
  beea1cfe