Variational Inference and Python
summary
Compares Markov Chain Monte Carlo to variational inference and demonstrates fitting a Bayesian neural network with PyMC3 and Edward, using ADVI and minibatch optimization; shows posterior predictive surfaces and per-point predictive uncertainty instead of full posterior samples; reports a held‑out test accuracy of 92.2% on a toy binary task; notes mean‑field approximations bias variance downward, optimization can hit local minima, and the method scales much faster though it remains approximate.
AI-written summary · about
chapters (8)
- 0:00Bayesian inference and PyMC3
- 4:14Markov chain Monte Carlo fundamentals
- 6:24MCMC pros and cons
- 10:42Variational inference overview
- 16:01Variational inference advantages
- 21:04Bayesian neural networks
- 27:38Interpretability and uncertainty
- 30:31Q&A
related talks
37mVariational Inference in Python
39mProbablistic Programming Data Science with PyMC3
39mProbablistic Programming Data Science with PyMC3
01h 29mA Beginner's Guide to Variational Inference
01h 31mProbabilistic Programming and Bayesian Inference in Python
44mAn introduction to PyMC3
12mBayesian Neural Networks With Variational Inference in PyTorch
21mProbabilistic Programming with Programmable Variational Inference