Mastering Gaussian Processes with PyMC
summary
Shows how to specify, fit and validate Gaussian processes in PyMC with concrete code and notebooks: HSGP approximation for speed, sparse inducing points for marginal sparse GPs, and latent GP likelihoods for non‑Gaussian data. Demonstrates on a 26,000‑row marathon example and an 827‑year cherry‑blossom temperature series, gives timings (MAP/opt and MCMC runs measured — ~4 minutes for a moderate binomial model, ~2 minutes for a smaller fit), and ships a public repo with runnable notebooks and diagnostics so the methods actually execute end‑to‑end.
AI-written summary · about
chapters (12)
- 0:00Gaussian Processes with PyMC intro
- 1:48Gaussian Processes for messy data
- 13:15GP fundamentals and kernels
- 27:40Implementing GP prior in PyMC
- 37:06Latent GP likelihood for non-normal
- 40:40Scaling GPs with large datasets
- 52:30PyMC case study intro
- 54:20Sports analytics model setup
- 1:00:06HSGP hyperparameters M and C
- 1:11:12Building PyMC model step-by-step
- 1:27:21Posterior analysis and diagnostics
- 1:30:12Sampling and validation
related talks
01h 30mA Primer on Gaussian Processes for Regression Analysis
01h 30mMaking Gaussian Processes Useful
44mActually using GPs in practice with PyMC
42mBayesian Non-parametric Models for Data Science using PyMC3
40mBdrygp: A Boundary-Integrated Gaussian Process Model For Computer Code Emulation
39mBayesian optimisation with scikit-learn
01h 23mProbabilistic Graphical Models, HMMs using PGMPY
31mBut what is a Gaussian process? Regression while knowing how certain you are