Blog

Deciding how much to hedge, and shaping forward curves.

Momentum strategies, turnover control and honest backtesting.

Running AI on phones and edge devices: compilation and quantisation.

Fast joint analyses with normalising flows.

Changing one word in a caption to edit an image with a diffusion model.

A Gaussian Process emulator for the power spectrum and its gradients.

Compressing weak lensing data with MOPED and emulating it with GPs.

Curated courses and lectures on machine learning and deep learning.

Highlights from ICLR, ICML, MLSS, OxML and GPSS, all online in 2020.

The steps I follow to take a research idea to a first-author paper.

Likelihood emulation with Gaussian Processes and Bayesian optimisation.

Andrew Ng at Imperial on narrow AI, small data, jobs and education.

Bayesian and machine learning methods for the next decade of cosmology.

Talks and tutorials from the Data Science Summer School in Paris.

Ten intensive days of machine learning, from k-NN to GANs.

Evidence for a Gaussian linear model, plus the Savage-Dickey ratio.

Bayesian evidence, Bayes factors and Occam's razor in model selection.

Highlights from the 2016 Bayesian School: KL divergence and BHMs.

Maximum likelihood fitting, applied to an A-Level physics experiment.

Three ways to sample any distribution, from SciPy to interpolation.

The Gibbs sampler, with a worked Bayesian regression example in Python.

Doing science directly on radio visibilities with Bayesian inference.

Lessons from PyConZA on culture, flexibility and growing communities.

Why hands-on JEDI workshops achieve more than conventional conferences.

How Gaussian Processes do regression, with and without noise.

What X-ray cavities in galaxy clusters reveal about AGN feedback.