This directory trains a model in the number of IP3s for a single volume, and across all IP3 concentrations.

It uses the data in the ml_training_data directory - visit that before this.

Run it in order following this guide.

Either:

• Pull the pre-loaded data using dvc pull, or
• Run the notebook cache_params0.nb to identify the ML parameters from the data, and write them to the cache folder.

Either:

• Pull the pre-loaded data using dvc pull, or
• Run the notebook cache_derivs0.nb to differentiate the ML parameters using total variation regularization, and write them to the cache folder.

Head to the train/training_data folder.

1. Create the graphs with the make_graph_rxns.nb notebook for the reaction-based model. To run this, first run the initialization cells in the funcs_network_rxns.nb to make the necessary function definitions.
2. Similarly, you can create the graphs for the parameter-based model with the corresponding methods.
3. Create the training data with the make_training_data.nb notebook.

The results are in the training_data_params and training_data_rxns folders.

Head to the train/train folder.

Either:

• Pull the pre-loaded trained networks using dvc pull, or
• Use the notebook train.nb to train the models.

The resulting trained networks and data are in the trained folder.

To calculate the Jacobians for the parameter transformations, see the transformation_jacobian folder.