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: 79.79%	auROC: 0.984	 auPRC: 0.420
Recall at 5% | 10% | 25% | 50% FDR: 0.0% | 0.0% | 11.2% | 40.8%	Num Positives: 4736	 Num Negatives: 966886


New best auPRC. Saving model.


Epoch 2:
Metrics across all datasets:
Balanced Accuracy: 81.12%	auROC: 0.985	 auPRC: 0.464
Recall at 5% | 10% | 25% | 50% FDR: 5.8% | 8.3% | 16.6% | 44.5%	Num Positives: 4736	 Num Negatives: 966886


New best auPRC. Saving model.


Epoch 3:
Metrics across all datasets:
Balanced Accuracy: 84.70%	auROC: 0.985	 auPRC: 0.466
Recall at 5% | 10% | 25% | 50% FDR: 5.4% | 7.4% | 16.9% | 45.1%	Num Positives: 4736	 Num Negatives: 966886


New best auPRC. Saving model.


Epoch 4:
Metrics across all datasets:
Balanced Accuracy: 86.76%	auROC: 0.981	 auPRC: 0.444
Recall at 5% | 10% | 25% | 50% FDR: 1.6% | 4.6% | 14.6% | 41.0%	Num Positives: 4736	 Num Negatives: 966886



Epoch 5:
Metrics across all datasets:
Balanced Accuracy: 82.65%	auROC: 0.980	 auPRC: 0.432
Recall at 5% | 10% | 25% | 50% FDR: 3.5% | 6.8% | 14.7% | 37.6%	Num Positives: 4736	 Num Negatives: 966886



Epoch 6:
Metrics across all datasets:
Balanced Accuracy: 80.62%	auROC: 0.978	 auPRC: 0.432
Recall at 5% | 10% | 25% | 50% FDR: 1.7% | 6.2% | 15.2% | 39.5%	Num Positives: 4736	 Num Negatives: 966886



Epoch 7:
Metrics across all datasets:
Balanced Accuracy: 81.64%	auROC: 0.978	 auPRC: 0.419
Recall at 5% | 10% | 25% | 50% FDR: 3.5% | 5.6% | 14.4% | 37.7%	Num Positives: 4736	 Num Negatives: 966886



Epoch 8:
Metrics across all datasets:
Balanced Accuracy: 81.38%	auROC: 0.976	 auPRC: 0.429
Recall at 5% | 10% | 25% | 50% FDR: 3.9% | 6.3% | 13.8% | 40.8%	Num Positives: 4736	 Num Negatives: 966886


Finished training after 8 epochs.
The best model's architecture and weights (from epoch 3) were saved to /users/marinovg/2018-01-25-histone-mods/tfdragonn-SequenceAndDnaseClassifier-K562-ATAC-cuts-YY1/model.arch.json and /users/marinovg/2018-01-25-histone-mods/tfdragonn-SequenceAndDnaseClassifier-K562-ATAC-cuts-YY1/model.weights.h5