dougbrn
commented
1 year ago TAPE has successfully run on superphot plus, for both pre-processing and classification using a trained model. This is a good first step, but we should consider workflows that have more active training requirements and how to best support them.
Machine Learning is a common tool to apply to time-domain data in astronomy. We should work to understand what supporting ML workflows could look like in TAPE, whether it's simply ensuring our ensemble data is servable to popular ML frameworks, or if more work is needed to make the ensemble a data structure that can support things like train/test splits, model storage and training/predictions, etc.