Skip to content
Snippets Groups Projects
Select Git revision
  • always-batch-training
  • dev-2D-ratio
  • dev-actmax
  • dev-adv
  • dev-balance-dataset
  • dev-friend
  • dev-genetic-algorithm
  • dev-input-transform-plotting
  • dev-mask
  • dev-memory
  • dev-organisation
  • dev-planing
  • dev-regression
  • dev-rnn
  • dev-score-plot
  • master default
16 results
You can move around the graph by using the arrow keys.
Created with Raphaël 2.2.014Dec135430Nov2928201915142130Oct29222117519Sep12104331Aug302928272624232221201716151413121098763231Jul3027242319Jun181413121187517May16159876543130Apr292726fix plot_input for usage with training batchesmastermasterfix WeightedRobustScalerMerge branch 'master' into dev-advdev-advdev-advadding option for leaky relumake adv learning rate tunablekeyboard interrupt for adversarial trainingrevert adam optimizer for adversarial to default lrapply class weights for adversarial targetsadding class_weight_target propertyallow arbitrary binning for decorr distnumber of decorr_bins as a parameterfixing loss plotadding BaseLogger and History callbacksensure fixed number of labels for adv targetname target layers in adversarial setupuse keras callbacks in adversarial trainingsome fixesAdversarial setup proof-of-conceptweights and layersMerge branch 'dev-regression'starting adversarial setup for decorrelationadd another hidden layer on top of regression targetsdev-regressiondev-regressionadjust compare.pyfixing evaluate_train_test for regression targetsremove duplicate functiontransform target for regressionweight individual lossesuse mse by default for regressionadd regression targets to modelsintroduce l_train/test to indicate the labels in case of output with regression targetsfill regression targets optionallyfix plot_input for inputs with a single valueadd option to plot input variables from batch generatoradd mechanism to set number of threads via environment variablelMerge remote-tracking branch 'origin/master'option to change default kernel initializerfixing evaluate_train_test (includes transformation now)option to ignore negative weight eventsimproved options for loss and ROC comparison plotsome comments