Learning Bayesian Statistics How to Choose & Use Priors, with Daniel Lee
Dec 20, 2023
Daniel Lee, an expert in Bayesian statistics, discusses the challenges in accurately depicting situations based on multiple data pieces. He explores identifying hidden objects, applying vision stats, building complex models in Bayesian statistics, and struggles in choosing priors.
AI Snips
Chapters
Transcript
Episode notes
Data Fusion Challenge
- Imagine an aircraft carrier surrounded by a dozen ships, each with radar.
- Slightly jittered radar data can lead to the false appearance of multiple threats if not fused correctly.
Model Complexity
- Build models complex enough to describe the phenomenon and explain the data.
- Consider measurement errors and incorporate them into the model for better fitting.
Bayesian Applicability
- Bayesian statistics are not a one-size-fits-all solution.
- Some problems, like fitting large language models (LLMs) with MCMC, may not be suitable for Bayesian methods.