@@ -6,9 +6,8 @@ The simple back-end of this package allows the user to create an MLP of a given
Media/data will be created in directories that begin with `_` by default - this is to follow the `.gitignore` rule that prevents the automatic pushing of updates to these files, which would slow commits down a lot
DO NOT RUN JUPYTER NOTEBOOK CODE AS DIRECTORIES ARE MESSED UP AT THE MOMENT - urgent fix
## TODO
- Need to put together optimisation code - choices to be made here: uniform sampling of hyperparameters vs. Bayesian optimisation, how to include speed/comp.cost component in optimisation (i.e. we don't want a huge lumbering network!).
- Diagnostics/tests need to be implemented more accessibly. We want to be able to evaluate the RMSE against a supplied dataset, along with RMSE/mismatch as a function of parameter space.
- The general ideas behind this approach need to be written up more formally, i.e. paper pre-pre-pre-draft general overview style.
