###### basset_train.lua
Train a convolutional neural network on the given data.
| Argument | Type | Description |
|---|---|---|
| data_file | HDF5 | Input training and validation data |
| Option | Description |
|---|---|
| -cuda | Run on GPGPU |
| -job | Table of job hyper-parameters |
| -max_epochs | Maximum training epochs to perform |
| -restart | Restart an interrupted training run using the given model file |
| -result | Write the loss value to this file (useful for Bayes Opt) |
| -save | Prefix for saved models [Default: dnacnn] |
| -seed | Seed the model with the parameters of another in the given model file |
| -rand | Random number generator seed |
| -stagnant_t | Allowed epochs with stagnant validation loss [Default: 10] |
###### basset_test.lua
Report model performance on the given test data, producing files with AUC and points along the ROC curves for each sample.
| Argument | Type | Description |
|---|---|---|
| model_file | Model | Saved model to use |
| data_file | HDF5 | Input training and validation data |
| out_dir | Output directory |
| Option | Description |
|---|---|
| -cuda | Run on GPGPU |
###### basset_predict.lua
Predict activity for a new set of sequences.
| Arguments | Type | Description |
|---|---|---|
| model_file | Model | Saved model to use |
| data_file | HDF5 | Input training and validation data |
| out_file | Output file |
| Option | Description |
|---|---|
| -cuda | Run on GPU [Default: False] |
| -norm | Normalize all targets to a 0.05 frequency |