while my_mcmc: gently(samples)
Do AIs suffer? On consciousness, ethics, and what respectful collaboration with AI systems might look like.
An intuitive guide to self-attention in GPT models: build up queries, keys and values step by step with plain NumPy.
Introducing PyMC Labs: a Bayesian consultancy built around the PyMC core team.
My PhD thesis on computational psychiatry: Bayesian modeling of decision-making in psychiatric disease.
My foreword to Bayesian Analysis with Python (2nd edition) by Osvaldo Martin.
Bayesian decision making in practice: optimize a supply chain under uncertainty by feeding PyMC posterior samples into a profit objective.
Hierarchical Bayesian neural networks in PyMC: share knowledge across related datasets with informative priors over network weights.
An intuitive, visual introduction to copulas: model correlated random variables with arbitrary marginal distributions in Python.
Random-walk Bayesian neural networks in PyMC: let network weights drift over time to handle non-stationary data.
Why hierarchical Bayesian models are awesome and tricky: divergences, the funnel of hell, and how the non-centered parameterization fixes them in PyMC.
Historical post: bridging PyMC3 and Lasagne to build Bayesian neural networks with variational inference on MNIST.
Bayesian neural networks in PyMC: fit a probabilistic neural net with ADVI variational inference and quantify prediction uncertainty.
MCMC sampling explained from scratch: code your own Metropolis-Hastings sampler in Python and build real intuition for how Bayesian inference works.
Getting started with data science in Python: a practical tour of pandas, seaborn and the Jupyter notebook ecosystem.
Hierarchical linear regression in PyMC: partial pooling and shrinkage explained hands-on with the classic radon dataset.
Historical post: distributing a parallel IPython notebook across a compute cluster with IPython.parallel.
Animating MCMC samplers in Python: watch Metropolis and friends explore a posterior with PyMC and Matplotlib.
Historical post: bolting the emcee ensemble sampler (the MCMC hammer) onto PyMC 2 models for parallel MCMC sampling in Python.
Robust Bayesian regression with Student-T likelihoods in PyMC and Bambi: how heavy tails tame outliers that wreck ordinary linear regression.
Bayesian GLMs in Python: build and fit a linear regression with PyMC and Bambi’s R-style formula syntax, and visualize posterior uncertainty.