@shaoyuawaw.bsky.social
I started programming in 2013, I still recall the atmosphere was types were for dinosaur companies, startup hackers use interpreted languages. Now, it feels like we have reversed...
October 31, 2025 at 2:01 AM
I started programming in 2013, I still recall the atmosphere was types were for dinosaur companies, startup hackers use interpreted languages. Now, it feels like we have reversed...
Great question, I imagine the goal of back-propagation should be to reach the core at the end of training? (not guaranteed?) Also, how can we come up with an interaction index using the core? By treating the core as a new grand coalition and apply Shapley's formula?
June 16, 2025 at 9:34 AM
Great question, I imagine the goal of back-propagation should be to reach the core at the end of training? (not guaranteed?) Also, how can we come up with an interaction index using the core? By treating the core as a new grand coalition and apply Shapley's formula?
Thanks for sharing this, which I think is the biggest revolution in statistical practice! I wrote down my learnings in fumin.github.io/notes/evalue/
Alarmingly, a recent survey[3] revealed that 56% of psychologists succumbed to the practice of deciding whether to collect more data after looking to see whether the results were significant. While collecting data until a desired p-value is obtained is obviously cheating, there are situations where the correct decision to be made is less clear: What if you submit a paper and the referee asks you to test a couple more subjects? Should you refuse because it invalidates your p-values!?
fumin.github.io
May 5, 2025 at 5:28 AM
Thanks for sharing this, which I think is the biggest revolution in statistical practice! I wrote down my learnings in fumin.github.io/notes/evalue/