GenomeFlowInterface Settings:
shuffle: True
pos_sampling_rate: 0.05
validation_chroms: ['chr22']
holdout_chroms: ['chr1', 'chr8', 'chr21']
optimizer: adam
learning rate: 0.0003
batch size: 256
epoch size: 2500000
max num of epochs: 100
early stopping metrics: auPRC
early stopping patience: 4

Epoch 1:
Metrics across all datasets:
Balanced Accuracy: 92.42%	auROC: 0.994	 auPRC: 0.356
Recall at 5% | 10% | 25% | 50% FDR: 0.1% | 0.1% | 0.1% | 30.6%	Num Positives: 1135	 Num Negatives: 1022650


New best auPRC. Saving model.


Epoch 2:
Metrics across all datasets:
Balanced Accuracy: 85.73%	auROC: 0.967	 auPRC: 0.389
Recall at 5% | 10% | 25% | 50% FDR: 0.9% | 1.0% | 1.3% | 38.1%	Num Positives: 1135	 Num Negatives: 1022650


New best auPRC. Saving model.


Epoch 3:
Metrics across all datasets:
Balanced Accuracy: 72.30%	auROC: 0.936	 auPRC: 0.340
Recall at 5% | 10% | 25% | 50% FDR: 0.1% | 0.1% | 1.1% | 34.0%	Num Positives: 1135	 Num Negatives: 1022650



Epoch 4:
Metrics across all datasets:
Balanced Accuracy: 88.55%	auROC: 0.990	 auPRC: 0.366
Recall at 5% | 10% | 25% | 50% FDR: 0.1% | 0.1% | 0.3% | 32.1%	Num Positives: 1135	 Num Negatives: 1022650



Epoch 5:
Metrics across all datasets:
Balanced Accuracy: 71.51%	auROC: 0.894	 auPRC: 0.338
Recall at 5% | 10% | 25% | 50% FDR: 0.3% | 0.3% | 2.6% | 36.7%	Num Positives: 1135	 Num Negatives: 1022650



Epoch 6:
Metrics across all datasets:
Balanced Accuracy: 94.14%	auROC: 0.986	 auPRC: 0.177
Recall at 5% | 10% | 25% | 50% FDR: 0.0% | 0.0% | 0.0% | 3.3%	Num Positives: 1135	 Num Negatives: 1022650



Epoch 7:
Metrics across all datasets:
Balanced Accuracy: 66.89%	auROC: 0.878	 auPRC: 0.276
Recall at 5% | 10% | 25% | 50% FDR: 0.2% | 0.2% | 0.2% | 24.3%	Num Positives: 1135	 Num Negatives: 1022650


Finished training after 7 epochs.
The best model's architecture and weights (from epoch 2) were saved to /users/marinovg/2018-01-25-histone-mods/tfdragonn-SequenceAndDNAseAnd4HistoneMarksClassifier_Separate-K562-DNAse-Duke-cuts-11HistoneMarks-SP1/model.arch.json and /users/marinovg/2018-01-25-histone-mods/tfdragonn-SequenceAndDNAseAnd4HistoneMarksClassifier_Separate-K562-DNAse-Duke-cuts-11HistoneMarks-SP1/model.weights.h5