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: 81.50%	auROC: 0.969	 auPRC: 0.078
Recall at 5% | 10% | 25% | 50% FDR: 0.0% | 0.0% | 0.0% | 0.0%	Num Positives: 1415	 Num Negatives: 996701


New best auPRC. Saving model.


Epoch 2:
Metrics across all datasets:
Balanced Accuracy: 86.96%	auROC: 0.983	 auPRC: 0.208
Recall at 5% | 10% | 25% | 50% FDR: 0.0% | 0.0% | 0.0% | 0.5%	Num Positives: 1415	 Num Negatives: 996701


New best auPRC. Saving model.


Epoch 3:
Metrics across all datasets:
Balanced Accuracy: 88.02%	auROC: 0.985	 auPRC: 0.283
Recall at 5% | 10% | 25% | 50% FDR: 0.0% | 0.0% | 0.0% | 21.8%	Num Positives: 1415	 Num Negatives: 996701


New best auPRC. Saving model.


Epoch 4:
Metrics across all datasets:
Balanced Accuracy: 87.51%	auROC: 0.986	 auPRC: 0.325
Recall at 5% | 10% | 25% | 50% FDR: 0.0% | 0.6% | 0.8% | 26.9%	Num Positives: 1415	 Num Negatives: 996701


New best auPRC. Saving model.


Epoch 5:
Metrics across all datasets:
Balanced Accuracy: 84.73%	auROC: 0.985	 auPRC: 0.357
Recall at 5% | 10% | 25% | 50% FDR: 0.6% | 0.7% | 0.8% | 32.9%	Num Positives: 1415	 Num Negatives: 996701


New best auPRC. Saving model.


Epoch 6:
Metrics across all datasets:
Balanced Accuracy: 86.82%	auROC: 0.988	 auPRC: 0.331
Recall at 5% | 10% | 25% | 50% FDR: 0.0% | 0.0% | 0.8% | 30.3%	Num Positives: 1415	 Num Negatives: 996701



Epoch 7:
Metrics across all datasets:
Balanced Accuracy: 91.09%	auROC: 0.989	 auPRC: 0.354
Recall at 5% | 10% | 25% | 50% FDR: 0.0% | 0.0% | 0.0% | 34.8%	Num Positives: 1415	 Num Negatives: 996701



Epoch 8:
Metrics across all datasets:
Balanced Accuracy: 90.40%	auROC: 0.989	 auPRC: 0.378
Recall at 5% | 10% | 25% | 50% FDR: 0.0% | 0.0% | 0.0% | 39.7%	Num Positives: 1415	 Num Negatives: 996701


New best auPRC. Saving model.


Epoch 9:
Metrics across all datasets:
Balanced Accuracy: 89.10%	auROC: 0.988	 auPRC: 0.388
Recall at 5% | 10% | 25% | 50% FDR: 0.8% | 0.8% | 0.8% | 41.7%	Num Positives: 1415	 Num Negatives: 996701


New best auPRC. Saving model.


Epoch 10:
Metrics across all datasets:
Balanced Accuracy: 87.45%	auROC: 0.988	 auPRC: 0.391
Recall at 5% | 10% | 25% | 50% FDR: 0.8% | 0.8% | 0.9% | 43.0%	Num Positives: 1415	 Num Negatives: 996701


New best auPRC. Saving model.


Epoch 11:
Metrics across all datasets:
Balanced Accuracy: 88.51%	auROC: 0.989	 auPRC: 0.394
Recall at 5% | 10% | 25% | 50% FDR: 0.0% | 0.0% | 0.8% | 44.0%	Num Positives: 1415	 Num Negatives: 996701


New best auPRC. Saving model.


Epoch 12:
Metrics across all datasets:
Balanced Accuracy: 88.37%	auROC: 0.989	 auPRC: 0.407
Recall at 5% | 10% | 25% | 50% FDR: 0.0% | 0.8% | 8.1% | 44.0%	Num Positives: 1415	 Num Negatives: 996701


New best auPRC. Saving model.


Epoch 13:
Metrics across all datasets:
Balanced Accuracy: 92.07%	auROC: 0.989	 auPRC: 0.400
Recall at 5% | 10% | 25% | 50% FDR: 0.0% | 0.0% | 0.0% | 43.3%	Num Positives: 1415	 Num Negatives: 996701



Epoch 14:
Metrics across all datasets:
Balanced Accuracy: 84.95%	auROC: 0.988	 auPRC: 0.410
Recall at 5% | 10% | 25% | 50% FDR: 1.0% | 1.0% | 13.0% | 43.1%	Num Positives: 1415	 Num Negatives: 996701


New best auPRC. Saving model.


Epoch 15:
Metrics across all datasets:
Balanced Accuracy: 86.56%	auROC: 0.989	 auPRC: 0.408
Recall at 5% | 10% | 25% | 50% FDR: 0.0% | 0.8% | 10.6% | 43.4%	Num Positives: 1415	 Num Negatives: 996701



Epoch 16:
Metrics across all datasets:
Balanced Accuracy: 91.11%	auROC: 0.990	 auPRC: 0.393
Recall at 5% | 10% | 25% | 50% FDR: 0.0% | 0.0% | 0.0% | 40.2%	Num Positives: 1415	 Num Negatives: 996701



Epoch 17:
Metrics across all datasets:
Balanced Accuracy: 90.63%	auROC: 0.989	 auPRC: 0.403
Recall at 5% | 10% | 25% | 50% FDR: 0.0% | 0.0% | 11.8% | 42.5%	Num Positives: 1415	 Num Negatives: 996701



Epoch 18:
Metrics across all datasets:
Balanced Accuracy: 91.11%	auROC: 0.989	 auPRC: 0.428
Recall at 5% | 10% | 25% | 50% FDR: 0.0% | 0.0% | 15.3% | 42.6%	Num Positives: 1415	 Num Negatives: 996701


New best auPRC. Saving model.


Epoch 19:
Metrics across all datasets:
Balanced Accuracy: 88.67%	auROC: 0.988	 auPRC: 0.425
Recall at 5% | 10% | 25% | 50% FDR: 0.0% | 0.0% | 14.3% | 43.3%	Num Positives: 1415	 Num Negatives: 996701



Epoch 20:
Metrics across all datasets:
Balanced Accuracy: 87.27%	auROC: 0.988	 auPRC: 0.423
Recall at 5% | 10% | 25% | 50% FDR: 0.0% | 0.0% | 12.8% | 43.5%	Num Positives: 1415	 Num Negatives: 996701



Epoch 21:
Metrics across all datasets:
Balanced Accuracy: 87.71%	auROC: 0.988	 auPRC: 0.406
Recall at 5% | 10% | 25% | 50% FDR: 0.0% | 0.0% | 8.4% | 42.9%	Num Positives: 1415	 Num Negatives: 996701



Epoch 22:
Metrics across all datasets:
Balanced Accuracy: 90.28%	auROC: 0.989	 auPRC: 0.400
Recall at 5% | 10% | 25% | 50% FDR: 0.0% | 0.0% | 8.1% | 42.3%	Num Positives: 1415	 Num Negatives: 996701



Epoch 23:
Metrics across all datasets:
Balanced Accuracy: 89.05%	auROC: 0.988	 auPRC: 0.423
Recall at 5% | 10% | 25% | 50% FDR: 0.0% | 0.0% | 12.3% | 44.2%	Num Positives: 1415	 Num Negatives: 996701


Finished training after 23 epochs.
The best model's architecture and weights (from epoch 18) were saved to /users/marinovg/2018-01-25-histone-mods/tfdragonn-SequenceClassifier-K562-CTCFL/model.arch.json and /users/marinovg/2018-01-25-histone-mods/tfdragonn-SequenceClassifier-K562-CTCFL/model.weights.h5