Commits · 4ea4173abd4625304048cb1fa9ae075d9566610e · Patrick Skillman-Lawrence / mhc_rank

Dec 10, 2017
- Refactor to simplify directory structure · 4ea4173a
  Tim O'Donnell authored 7 years ago
  
  4ea4173a
- formatting changes · 6f6070b2
  Tim O'Donnell authored 7 years ago
  
  6f6070b2
Dec 04, 2017
- Even faster peptide encoding · b2b032c7
  Tim O'Donnell authored 7 years ago
  
  b2b032c7
- Fix predict test · 4e68da43
  Tim O'Donnell authored 7 years ago
  
  4e68da43
- Address code review comments · 0f9394e2
  Tim O'Donnell authored 7 years ago
  
  0f9394e2
- Train on alleles with >=75 measurements (decreased from 100) · 0ae107b1
  Tim O'Donnell authored 7 years ago
  
  0ae107b1
- Update downloads to include new models · 6dfa6417
  Tim O'Donnell authored 7 years ago
  
  6dfa6417
Dec 01, 2017
- Switch back to dense=16 · 28041d1c
  Tim O'Donnell authored 7 years ago
  
  28041d1c
- bump mhcflurry version · bdc672f4
  Tim O'Donnell authored 7 years ago
  
  pre-1.0
  
  bdc672f4
- fix requirements · 1f38fa81
  Tim O'Donnell authored 7 years ago
  
  1f38fa81
- lazy load keras · dd3e61e2
  Tim O'Donnell authored 7 years ago
  
  dd3e61e2
- 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