madhavajay
commented
9 months ago @jokikim we do have something that might satisfy your desire for dashboards on the roadmap.
As for doing things like Scikit learn that should be possible in the new 0.8.4 custom container workload APIs. The release will be in the next few days but the latest beta should be just as stable.
https://github.com/OpenMined/PySyft/blob/dev/notebooks/api/0.8/10-container-images.ipynb
Question
Based on PySyft's underlying technology, do you believe that 1) there's a way to allow more robust analytics, perhaps the ability to create dashboards and reporting based on the private data, and 2) a built-in way to run classical ML models, such as through Scikit-Learn API?
Further Information
I have two questions, for which the contexts are below:
1) Dashboarding: I am not well-versed in Differential Privacy/Federated Learning and other technologies that drive PySyft, but I'm wondering what the full potential of PySyft could look like, beyond simply running DL models. Specifically, I wonder if there's a way, or could ever be a way, to build descriptive analytics around mock/private data, and potentially build Dashboards and Reports around them. I understand that the Data Scientists may not have the actual column names and such, but would there be a way to maintain the underlying relationships between the data so that they could report the relationships between features?
2) Classical ML: I understand that PySyft data is represented as tensors, but does PySyft have any intention on building in Classical ML models, such as Random Forest etc., as well? Is there a reason why Classical ML hasn't been incorporated much to this point?
Screenshots
System Information
Additional Context