Understanding the transition from Phase 1 and Phase 2

This question is two-folds, technical and logistical.

Technical:
Would you have any information about which datasets are most likely to be federated in phase-2?
Obviously, the decentralizedd/federated systems need to send some information back to the centralized system to ultimately make a decision. Is there an idea for how much data is acceptable (if we want to do some architectures such as data dimensionality reduction thru PCA/autoencoders, this would provide a good idea for how many dimensions we want to reduce).

Logistical:
How many “finalist” teams will you be sending to phase 2? This gives us an idea for our cost-benefit analysis in terms of how much we want to design our system for federate system to set us up to compete in phase-2, as compared to focusing on making a good model
Also, how much support can we expect in phase-2 in terms of help from NASA (are they just giving us their tools to use for federated training, or should we expect subject matter expert support)?

Appreciate it!

1 Like

Hi @pennswood Glad you’re joining us for another airport challenge! We plan to release more details about Phase 2 after the launch of the Prescreened Arena tomorrow. In the meantime, I will check in with the NASA team to see if there is any detail we can provide about your specific questions.

1 Like

Relevant to your question: we added some more context to the problem description. Head over to the newly updated Phase 2 section of the Problem Description to find out more, including which features will be private and which will be public to your federated model. We hope this helps you get started thinking about how to federate your winning model!

1 Like

Thank you. This is very helpful for us to determine our infrastructure moving forward!