This is a collection of ideas for future directions of this project, mainly intended to keep track of suggestions. The ideas may ideally be executed through funding channels such as GSOC
- Provide alternate penalty functions e.g. Tikhonov regularizers for the L2 part of the loss function: very useful for time series models
- Implement group lasso/ group elastic net: very useful to regularize in groups
- Implement SGD variants by potentially integrating with
sklearn
: useful for big data - Make a contribution to Spark's MLLib: useful for streaming data
- Interface with companion packages such as proposed
Spykes
andNeuroSTRF
to demonstrate use cases for neural data - Interface with big functional neural data as Jeremy Freeman's mouse or zerbrafish datasets.