Commits · d6050ad5ec16700f3b4690a0374fae288c77cd23 · Patrick Skillman-Lawrence / mhc_rank

Dec 01, 2017
- Bump to Keras==2.1.2 · d6050ad5
  Tim O'Donnell authored 7 years ago
  
  d6050ad5
- Fix encoding for 0-length sequences · 4bb0738b
  Tim O'Donnell authored 7 years ago
  
  4bb0738b
Nov 30, 2017
- More speed efficient variable-length peptide encoding · 873a60a5
  Tim O'Donnell authored 7 years ago
  
  873a60a5
- Switch back to dense layer size 32 from 16 · d65f7740
  Tim O'Donnell authored 7 years ago
  
  d65f7740
Nov 28, 2017
- fix · bdee8330
  Tim O'Donnell authored 7 years ago
  
  bdee8330
- update speed benchmark · 1d75ca19
  Tim O'Donnell authored 7 years ago
  
  1d75ca19
- fix · 7d323a86
  Tim O'Donnell authored 7 years ago
  
  7d323a86
- Add time remaining estimation to training script · f179489b
  Tim O'Donnell authored 7 years ago
  
  f179489b
- update models_class1_experiments1 to use minibatch_size=128 · 8f7fc5b1
  Tim O'Donnell authored 7 years ago
  
  8f7fc5b1
- Free up memory: get rid of train data as models are trained · 672bf7fb
  Tim O'Donnell authored 7 years ago
  
  672bf7fb
- better progress printing · e062613c
  Tim O'Donnell authored 7 years ago
  
  e062613c
- Print progress no more often than once every 5 seconds while training · b1dbb362
  Tim O'Donnell authored 7 years ago
  
  b1dbb362
- Clearer output during training · c0fbebe1
  Tim O'Donnell authored 7 years ago
  
  c0fbebe1
- Experimental support for parallelization (Theano only) in train_allele_specific_models_command · 5b4dfbd2
  Tim O'Donnell authored 7 years ago
  
  5b4dfbd2
- implement Class1AffinityPredictor.merge() method · c6990666
  Tim O'Donnell authored 7 years ago
  
  c6990666
- bump pandas requirement · 2f02695d
  Tim O'Donnell authored 7 years ago
  
  2f02695d
- Use 1M peptides for calibration · ef9a61a9
  Tim O'Donnell authored 7 years ago
  
  ef9a61a9
- arg tweak · ffcb693f
  Tim O'Donnell authored 7 years ago
  
  ffcb693f
- Add minibatch_size hyperparameter · 33b1b5d4
  Tim O'Donnell authored 7 years ago
  
  33b1b5d4
- Change prediction batch_size from 32 to 4096 for better performance · 38adfb4d
  Tim O'Donnell authored 7 years ago
  
  38adfb4d
Nov 27, 2017
- peptide encoding optimization · cbb649d6
  Tim O'Donnell authored 7 years ago
  
  cbb649d6
- Much faster performance (~5X improvement) in amino acid encoding · 41985338
  Tim O'Donnell authored 7 years ago
  
  41985338
Nov 26, 2017
- fix · 78cdb265
  Tim O'Donnell authored 7 years ago
  
  78cdb265
- more efficient peptide encoding in calibrate_percentile_ranks · 1ca484c7
  Tim O'Donnell authored 7 years ago
  
  1ca484c7
- update · a2a11cae
  Tim O'Donnell authored 7 years ago
  
  a2a11cae
- Switch from 32 hidden layer size to 16 in standard models · ee242d39
  Tim O'Donnell authored 7 years ago
  
  ee242d39
Nov 25, 2017
- Add cross-validation GENERATE.sh and related scripts · 55e3cfc1
  Tim O'Donnell authored 7 years ago
  
  55e3cfc1
- add widelocal variant · 700535fa
  Tim O'Donnell authored 7 years ago
  
  700535fa
- fix · f21d9ebb
  Tim O'Donnell authored 7 years ago
  
  f21d9ebb
- update models_class1_experiments1 to include embedding and onehot input variations · ec543a8a
  Tim O'Donnell authored 7 years ago
  
  ec543a8a
- update models_class1_experiments1 · 13c509e2
  Tim O'Donnell authored 7 years ago
  
  13c509e2
- json -> yaml renames · 0cfd7e0d
  Tim O'Donnell authored 7 years ago
  
  0cfd7e0d
- new architectures · f91a85cf
  Tim O'Donnell authored 7 years ago
  
  f91a85cf
- Fix tests, add logging to calibrate_percentile_ranks, add SIGUSR1 trick to train command · ac3f874f
  Tim O'Donnell authored 7 years ago
  
  ac3f874f
- README update (mhctools mention) and percentile rank tweaks · 72362eb2
  Tim O'Donnell authored 7 years ago
  
  72362eb2
Nov 24, 2017
- fix · 2b2ce0ce
  Tim O'Donnell authored 7 years ago
  
  2b2ce0ce
- Add throw argument to percentile_ranks · e4c21bcb
  Tim O'Donnell authored 7 years ago
  
  e4c21bcb
Nov 15, 2017
- tweak · 868caf01
  Tim O'Donnell authored 7 years ago
  
  868caf01
- comment tweak · 4fbdd62c
  Tim O'Donnell authored 7 years ago
  
  4fbdd62c
- Support percentile ranks · 99e74eeb
  Tim O'Donnell authored 7 years ago
  
  * First cut of supporting quantiles ("percent ranks") per allele. Closes #87. * Some refactoring of amino_acid.py This builds on #114
  99e74eeb