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: 81.93%	auROC: 0.958	 auPRC: 0.249
Recall at 5% | 10% | 25% | 50% FDR: 9.1% | 12.1% | 16.5% | 21.0%	Num Positives: 1401	 Num Negatives: 997988


New best auPRC. Saving model.


Epoch 2:
Metrics across all datasets:
Balanced Accuracy: 93.94%	auROC: 0.992	 auPRC: 0.438
Recall at 5% | 10% | 25% | 50% FDR: 13.7% | 14.6% | 18.5% | 36.2%	Num Positives: 1401	 Num Negatives: 997988


New best auPRC. Saving model.


Epoch 3:
Metrics across all datasets:
Balanced Accuracy: 94.06%	auROC: 0.988	 auPRC: 0.349
Recall at 5% | 10% | 25% | 50% FDR: 4.4% | 8.1% | 12.8% | 20.9%	Num Positives: 1401	 Num Negatives: 997988



Epoch 4:
Metrics across all datasets:
Balanced Accuracy: 85.70%	auROC: 0.981	 auPRC: 0.458
Recall at 5% | 10% | 25% | 50% FDR: 11.3% | 13.8% | 21.2% | 40.1%	Num Positives: 1401	 Num Negatives: 997988


New best auPRC. Saving model.


Epoch 5:
Metrics across all datasets:
Balanced Accuracy: 73.91%	auROC: 0.961	 auPRC: 0.359
Recall at 5% | 10% | 25% | 50% FDR: 12.6% | 15.5% | 19.5% | 25.7%	Num Positives: 1401	 Num Negatives: 997988



Epoch 6:
Metrics across all datasets:
Balanced Accuracy: 83.91%	auROC: 0.977	 auPRC: 0.400
Recall at 5% | 10% | 25% | 50% FDR: 7.9% | 11.0% | 15.6% | 29.2%	Num Positives: 1401	 Num Negatives: 997988



Epoch 7:
Metrics across all datasets:
Balanced Accuracy: 85.83%	auROC: 0.972	 auPRC: 0.416
Recall at 5% | 10% | 25% | 50% FDR: 4.7% | 6.6% | 11.8% | 36.4%	Num Positives: 1401	 Num Negatives: 997988



Epoch 8:
Metrics across all datasets:
Balanced Accuracy: 77.29%	auROC: 0.936	 auPRC: 0.327
Recall at 5% | 10% | 25% | 50% FDR: 6.1% | 6.6% | 7.6% | 24.2%	Num Positives: 1401	 Num Negatives: 997988



Epoch 9:
Metrics across all datasets:
Balanced Accuracy: 92.13%	auROC: 0.987	 auPRC: 0.418
Recall at 5% | 10% | 25% | 50% FDR: 12.6% | 16.0% | 21.1% | 31.3%	Num Positives: 1401	 Num Negatives: 997988


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-SequenceAndDNAseAnd6HistoneMarksClassifier_Separate-K562-DNAse-Duke-cuts-11HistoneMarks-REST/model.arch.json and /users/marinovg/2018-01-25-histone-mods/tfdragonn-SequenceAndDNAseAnd6HistoneMarksClassifier_Separate-K562-DNAse-Duke-cuts-11HistoneMarks-REST/model.weights.h5