#166 PTGP: A New Gaussian Process Library, with Bill Engels & Jesse Grabowski

<p><b><strong>Support &amp; Resources</strong></b><br />→ Support the show on <a href="https://www.patreon.com/c/learnbayesstats" rel="noopener noreferrer nofollow" target="_blank">Patreon</a><br />→ <a href="https://topmate.io/alex_andorra/1011122" rel="noopener noreferrer nofollow" target="_blank">Bayesian Modeling Course</a> (first 2 lessons free)<br /><br />Our theme music is « Good Bayesian », by Baba Brinkman (feat MC Lars and Mega Ran). Check out his<a href="https://bababrinkman.com/" rel="noopener noreferrer nofollow" target="_blank"> awesome work!</a></p><p><b><strong>Takeaways</strong></b>:</p><p>Q: What is PTGP and why did Bill and Jesse build it?<br />A: PTGP is a new Gaussian process library built on PyTensor and PyMC, aimed squarely at practitioners rather than researchers assembling their own GP methods from papers. Bill Engels, who wrote PyMC's original GP submodule during a Google Summer of Code, built it because most GP libraries implement one method well but give little guidance on when to use it or how to fix it when it breaks. PTGP is opinionated by design: it picks battle-tested algorithms, tells you which one fits your data size, and ships the debugging knowledge alongside the code.<br /><br />Q: Why are Gaussian processes described as sitting at the intersection of statistics and machine learning?<br />A: A Gaussian process starts from a simple idea: things that are close in the input space should be close in the output space, and the kernel function defines exactly what "close" means for your problem. That makes GPs look like machine learning (let the data speak through a flexible function) while staying fully interpretable once you've chosen a kernel, since observing data collapses the process into an ordinary multivariate normal.</p><p><b><strong>Full takeaways </strong></b><a href="https://learnbayesstats.com/episode/gaussian-processes-kernels-pytensor-ptgp-bill-engels-jesse-grabowski" rel="noopener noreferrer nofollow"><b><strong>here</strong></b></a><b><strong>!</strong></b><br /><br /><b><strong>Chapters</strong></b>:<br /><br />05:51 Where do Gaussian processes actually get used in practice?<br />09:07 What do a kernel's length scale and amplitude actually control?<br />11:58 What is HSGP and how does it combine with hierarchical models?14:51 Where does PTGP fit in a modern data science workflow?<br />21:01 Are Gaussian processes interpretable, or are they a black box?27:49 Why aren't Gaussian processes used everywhere alrea