I guess a welcome is in order…#
Kind of you to drop by! This is the personal website of Olivier Ma.
He conocido lo que ignoran los griegos: la incertidumbre. (La lotería en Babilonia, Jorge Luis Borges)
The name of the blog, Lottery of Babylonian variations, is an (intended) variation of the title of a short story from Borges, The lottery in Babylon. As Borges and Deleuze often argued, life is all about repetition and variation. And the same is certainly true for probabilities. It’s a story about chance, uncertainty, and a mechanism that started as a fool’s play, but evolved, little by little, into a society, a religion, and ultimately the whole universe. It seemed to be quite well suited here.
This blog is mostly about deep learning and statistical modeling, especially from a Bayesian perspective, so that’s likely what you are going to find here.
But you may also find some of my other sundry ruminations here.
On this blog I write in Chinese, which is my native language; and English, which I’m reasonably comfortable with; and French, which I’m reasonably uncomfortable with. I’m also learning Spanish and German.
Contact me at olivier.ma.dq@gmail.com.
I took the Solve It with Code course from answer.ai, and Advent of Code is used as practice.
I always thought my Python skills were quite mediocre; I always used to in a “just get things done” style, so this is also a chance to improve my Python. For this reason I want to mostly stick with the standard packages and pythonic styles, and only use extra packages when it’s too cumbersome otherwise.
...
This is an introduction to NumPyro, using the Eight Schools model as example.
Here we demonstrate the effects of model reparameterisation. Reparameterisation is especially important in hierarchical models, where the joint density tend to have high curvatures.
1 2 3 4 5 6 7 8 import numpy as np import jax.numpy as jnp import numpyro import numpyro.distributions as dist from jax import random from numpyro.infer import MCMC, NUTS, Predictive rng_key = random.PRNGKey(0) Here we are using the classic eight schools dataset from Gelman et al. We have collected the test score statistics for eight schools, including the mean and standard error. The goal, is to determine whether some schools have done better than others. Note that since we are working with the mean and standard error of eight different schools, we are actually modeling the statistical analysis resutls of some other people: this is essentially a meta analysis problem.
...