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: 87.03%	auROC: 0.986	 auPRC: 0.249
Recall at 5% | 10% | 25% | 50% FDR: 0.1% | 0.1% | 2.1% | 19.8%	Num Positives: 1785	 Num Negatives: 1012540


New best auPRC. Saving model.


Epoch 2:
Metrics across all datasets:
Balanced Accuracy: 91.52%	auROC: 0.991	 auPRC: 0.415
Recall at 5% | 10% | 25% | 50% FDR: 4.6% | 6.6% | 17.1% | 38.9%	Num Positives: 1785	 Num Negatives: 1012540


New best auPRC. Saving model.


Epoch 3:
Metrics across all datasets:
Balanced Accuracy: 91.98%	auROC: 0.992	 auPRC: 0.485
Recall at 5% | 10% | 25% | 50% FDR: 12.2% | 14.4% | 26.7% | 48.8%	Num Positives: 1785	 Num Negatives: 1012540


New best auPRC. Saving model.


Epoch 4:
Metrics across all datasets:
Balanced Accuracy: 89.80%	auROC: 0.993	 auPRC: 0.491
Recall at 5% | 10% | 25% | 50% FDR: 7.8% | 13.8% | 27.6% | 49.6%	Num Positives: 1785	 Num Negatives: 1012540


New best auPRC. Saving model.


Epoch 5:
Metrics across all datasets:
Balanced Accuracy: 87.38%	auROC: 0.993	 auPRC: 0.429
Recall at 5% | 10% | 25% | 50% FDR: 0.0% | 0.0% | 20.4% | 42.0%	Num Positives: 1785	 Num Negatives: 1012540



Epoch 6:
Metrics across all datasets:
Balanced Accuracy: 88.62%	auROC: 0.993	 auPRC: 0.458
Recall at 5% | 10% | 25% | 50% FDR: 0.2% | 2.5% | 23.2% | 47.3%	Num Positives: 1785	 Num Negatives: 1012540



Epoch 7:
Metrics across all datasets:
Balanced Accuracy: 91.39%	auROC: 0.993	 auPRC: 0.452
Recall at 5% | 10% | 25% | 50% FDR: 0.0% | 0.0% | 22.5% | 45.7%	Num Positives: 1785	 Num Negatives: 1012540



Epoch 8:
Metrics across all datasets:
Balanced Accuracy: 83.36%	auROC: 0.991	 auPRC: 0.430
Recall at 5% | 10% | 25% | 50% FDR: 0.0% | 0.0% | 21.5% | 42.6%	Num Positives: 1785	 Num Negatives: 1012540



Epoch 9:
Metrics across all datasets:
Balanced Accuracy: 87.71%	auROC: 0.993	 auPRC: 0.468
Recall at 5% | 10% | 25% | 50% FDR: 2.9% | 5.8% | 18.6% | 49.2%	Num Positives: 1785	 Num Negatives: 1012540


Finished training after 9 epochs.
The best model's architecture and weights (from epoch 4) were saved to /users/marinovg/2018-01-25-histone-mods/tfdragonn-SequenceAndDnaseClassifier-K562-DNAse-Duke-cuts-GATA2/model.arch.json and /users/marinovg/2018-01-25-histone-mods/tfdragonn-SequenceAndDnaseClassifier-K562-DNAse-Duke-cuts-GATA2/model.weights.h5