Data Collection is Hard. You Should Try It.
No, seriously: collecting data is a good idea, even for selfish reasons!
No, seriously: collecting data is a good idea, even for selfish reasons!
WebSockets with the Tornado web framework is a simple, robust way to handle streaming data. I walk through a minimal example and discuss why these tools are good for the job.
An exciting professional update: I’ve joined Flatiron Health as a data scientist!
cryptics.eigenfoo.xyz
— A Dataset of Cryptic Crossword Clues
Permalink
cryptics.eigenfoo.xyz is a dataset of cryptic crossword clues, collected from various blogs and publicly available digital archives.
In this blog post, we’ll break down what tensor computation libraries actually are, and how they differ. We’ll take a detailed look at some popular libraries, and end with an observation on the future of Theano in the context o...
littlemcmc
— A Standalone HMC and NUTS Sampler in Python
Introducing littlemcmc — a lightweight and performant implementation of HMC and NUTS in Python, spun out of the PyMC project.
Floating-point format is not a crucial consideration in deep learning, but it can make a significant difference. What is floating-point, why should you (a deep learning practictioner) care, and what can you do about it?
I’ve recently had to learn a lot about natural language processing — specifically, Transformer-based models. I thought I’d write up my reading and research and post it.
I explored the possibility of simplifying PyMC4’s model specification API by manipulating the Python abstract syntax tree (AST) of the model code.
I benchmarked various mass matrix adaptation methods in PyMC3. Sane defaults are easy to take for granted: it’s more nuanced than I initially expected!
stan-vim
A Vim plugin for Stan, offering syntax highlighting, automatic indentation and code folding. Check it out!
In this blog post, we’ll break down what probabilistic programming frameworks are made up of, and how the various pieces are organized and structured. We’ll take a look at some open source frameworks as examples.
Some exciting personal news: I’ve graduated from The Cooper Union and I’m joining Point72 Asset Management as a data scientist/research analyst!
I ported Jonas Lindeløv’s post, Common Statistical Tests are Linear Models from R to Python. Check it out on my blog, GitHub, or Binder!
In this blog post, we’ll take a look at two extensions to the multi-armed bandit. The first allows the bandit to model nonstationary rewards distributions, whereas the second allows the bandit to model context.
My current project involves working with a class of fairly niche and interesting neural networks that aren’t usually seen on a first pass through deep learning. I thought I’d write up my reading and research and post it.
An annotated reading list on modern computational methods for Bayesian inference — Markov chain Monte Carlo (MCMC), variational inference (VI) and some other (more experimental) methods.
A talk I gave about a recent project to model hate speech on Reddit. In this blog post, I describe the thought processes behind the project, and the stumbling blocks encountered along the way.
This blog post summarizes some literature on probabilistic and Bayesian matrix factorization methods, keeping an eye out for applications to one specific task in NLP: text clustering.
In this blog post I hope to show that there is more to Bayesianism than just MCMC sampling and suffering, by demonstrating a Bayesian approach to a classic reinforcement learning problem: the multi-armed bandit.
This is a compilation of notes, tips, tricks and recipes for Bayesian modelling that I’ve collected from everywhere: papers, documentation, peppering my more experienced colleagues with questions.
A recent project on trying to model hate speech on Reddit through text clustering — from ‘nimble navigators’ to ‘swamp creatures’, ‘spezzes’ to the ‘Trumpire’.
Latent Dirichlet allocation is a well-known and popular model in machine learning and natural language processing, but it really sucks sometimes. Here’s why.
What is learning using privileged information (LUPI), how do I do it, and why should I care? A brief introduction to LUPI and SVM+.
Everything that you wanted to know (and more!) about linear discriminant analysis (LDA) — how it works, why it works, and how to use it.
I interned at Quantopian this summer, where I contributed some exciting stuff to their open-source portfolio analytics engine, pyfolio. Check it out!
This is the first post of what will (hopefully) be a cool and interesting blog. Hope you like it!