#DataDistribution
A key theme: why isn't BitTorrent used more for large datasets like OSM, container registries, or package repos? Negative perception, complex firewall rules, and lack of built-in corporate clients are major blockers. #DataDistribution 2/7
September 23, 2025 at 7:00 PM
9/14 godelski: Test data should belong to the same target distribution BUT be sufficiently distant from your training data to measure generalization. 🎯 #Generalization #MachineLearning #DataDistribution
May 2, 2025 at 10:10 AM