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: 91.89%	auROC: 0.995	 auPRC: 0.321
Recall at 5% | 10% | 25% | 50% FDR: 1.1% | 3.8% | 10.0% | 26.5%	Num Positives: 548	 Num Negatives: 1024331


New best auPRC. Saving model.


Epoch 2:
Metrics across all datasets:
Balanced Accuracy: 89.78%	auROC: 0.990	 auPRC: 0.257
Recall at 5% | 10% | 25% | 50% FDR: 0.5% | 0.5% | 1.5% | 18.6%	Num Positives: 548	 Num Negatives: 1024331



Epoch 3:
Metrics across all datasets:
Balanced Accuracy: 89.60%	auROC: 0.983	 auPRC: 0.203
Recall at 5% | 10% | 25% | 50% FDR: 0.0% | 0.0% | 0.0% | 14.2%	Num Positives: 548	 Num Negatives: 1024331



Epoch 4:
Metrics across all datasets:
Balanced Accuracy: 80.28%	auROC: 0.981	 auPRC: 0.252
Recall at 5% | 10% | 25% | 50% FDR: 0.5% | 0.5% | 1.1% | 19.9%	Num Positives: 548	 Num Negatives: 1024331



Epoch 5:
Metrics across all datasets:
Balanced Accuracy: 84.49%	auROC: 0.989	 auPRC: 0.236
Recall at 5% | 10% | 25% | 50% FDR: 1.6% | 2.0% | 3.1% | 9.3%	Num Positives: 548	 Num Negatives: 1024331



Epoch 6:
Metrics across all datasets:
Balanced Accuracy: 80.79%	auROC: 0.987	 auPRC: 0.272
Recall at 5% | 10% | 25% | 50% FDR: 2.0% | 3.3% | 6.6% | 21.2%	Num Positives: 548	 Num Negatives: 1024331


Finished training after 6 epochs.
The best model's architecture and weights (from epoch 1) were saved to /users/marinovg/2018-01-25-histone-mods/tfdragonn-SequenceAndDnaseClassifier-K562-ATAC-02-cuts-CTCF-N-ChIP/model.arch.json and /users/marinovg/2018-01-25-histone-mods/tfdragonn-SequenceAndDnaseClassifier-K562-ATAC-02-cuts-CTCF-N-ChIP/model.weights.h5