All Posts

Getting value out of coding agents for low-urgency, low-impact projects.

I’ve been a somewhat late adopter to coding agents: until ~1 month ago I was mainly using cursor and asking it to suggest local edits, or occasionally bigger refactors. In…
May 4, 2026

Write research papers to be skimmed

Don’t write your paper for somebody to read it, write for somebody to skim it.
May 4, 2026

First time encountering a hallucinated citation of my paper.

I recently ran an LLM deep research query about large matrices of Tanimoto coefficients,1 and one of its references was the paper “Bandwidth adjustable Tanimoto kernel: a…
May 1, 2026

Book review: The Drug Hunters

Mar 11, 2026

Behaviour-based evaluations in Bayesian optimization

I think the Bayesian optimization (BO) research community needs to change its evaluation practices. In a previous post I explained how I think users should generally bring…
Feb 21, 2026

Rebranding BO away from “black-box” and towards “model-based”

In a recent blog post (link) I described my “model-centric” view of Bayesian optimization (BO), essentially arguing that the model is the most important component of BO and…
Feb 16, 2026

Clarifying noise vs model misspecification in Gaussian Process models (and its importance in BO)

The most common kind of Gaussian process (GP) model is:
Feb 14, 2026

Tyler Cowen’s book on talent

Cross-posting my summary of the book “Talent: How to Identify Energizers, Creatives, and Winners Around the World” from my GitHub: link.
Feb 13, 2026

We are underselling the modularity of Bayesian optimization

I think the Bayesian optimization (BO) community is vastly underselling one of BO’s most practically appealing features: modularity. A “modular” algorithm is one composed of…
Feb 9, 2026

Why I don’t care about toy benchmarks in BO

A lot of Bayesian optimization (BO) papers include experiments on “toy” functions. Examples of such toy functions are Ackley, Rosenbrock, Hartmann, and Branin. I basically…
Feb 8, 2026

We have forgotten about utility functions in BO (whoops!)

Bayesian decision theory is one of the best justifications for BO- particularly for myopic acquisition functions like expected improvement. However, these acquisition…
Jan 27, 2026

My model-centric view of Bayesian optimization

A lot of conversations at NeurIPS this year made me think that I view the role of the surrogate model in Bayesian optimization a bit differently than many other researchers…
Jan 25, 2026

Predictions for ML/AI in 2026 (and 2025 predictions re-visited).

Last year I made a bunch of predictions about ML, and since 2025 is over it’s time to grade myself and repeat this exercise.
Jan 5, 2026

New Year’s Resolutions for 2026 (and scoring 2025 resolutions)

Happy 2026! Here is my self-assessment on last year’s goals and my goals for this year (2026).
Jan 5, 2026

Some odd requests I’ve received to put links on my website.

In the ~8 years since I started my website I’ve had a bunch of people reach out to me requesting that a link be added to my “links” page (currently called resources). The…
Jan 4, 2026

Thoughts on writing PhD reference letters.

2025 was the first year where I wrote reference letters for research students who I supervised. In this post I will explain my strategy towards writing letters and how…
Dec 30, 2025

Review of NeurIPS 2025

I added the NeurIPS workshops this year (not the main conference).1 Here are my takeaways. I’ll follow up with predictions for 2026 in a later post (EDIT: here is that post).
Dec 18, 2025

Why listen to podcasts?

Although everybody seems to listen to podcasts these days, some people do not, and I’ve found myself explaining several times why I choose to listen to them (as opposed to…
Nov 16, 2025

Finally, an ML conference review guide!

ICLR 2026 released a detailed review guide as part of its review process (link). Let’s analyze it!
Nov 4, 2025

Justifying expected utility maximization from first principles (Von Neumann–Morgenstern).

I’ve recently found myself arguing for expected utility maximization as an approach to practical decision making problems at work (mostly regarding molecule selection). This…
Nov 2, 2025

Latent Space COWBOYS: a VAE-BO method I can actually buy into!

I am generally pessimistic about BO (Bayesian optimization) methods which use VAE embeddings as part of the model. Mostly this is because distance in a VAE’s latent space…
Oct 5, 2025

An accessible introduction to mutual information using a d20

Scientists want to do ‘informative’ experiments. What makes an experiment informative? I believe that information theory has a lot of answers. In this post I will try to explain information theory’s d…

Aug 10, 2025

The case for public Slack channels only (no DMs).

Working at a company with a distributed team and hybrid work, a lot of work gets done on Slack. Slack can be very noisy, and I used to wish that fewer people would post in…
Jul 27, 2025

Tornados are just wind at close range.

I find tornados really interesting: dark funnels of power and destruction. From far away they look like a “wall of cloud”, and over the last 6 months I spent a bunch of time…
Jul 26, 2025

Time zones: why you shouldn’t use the terms EST or GMT

This post is about a small mistake in how people discuss time zones which I find mildly annoying: using the term EST (Eastern Standard Time) to refer to the current time in…
Jul 13, 2025

The hardest bugs are non-bugs

Debugging computer programs is hard.
Jul 12, 2025

Blogging in the LLM age: a second golden age for blogs?

LLMs (large language models) are currently scraping all text on the public internet, training on it, and spitting out variants of that text in response to queries. I think…
Jun 23, 2025

My review guide for machine learning conference papers

There is no official step by step guide for how to review ML conference papers at venues like NeurIPS/ICML/ICLR.1 In this post, I try to explain my guide. It is not…
Jun 22, 2025

Problems with ML for toxicity prediction.

I recently read a review paper by Seal et al about machine learning for toxicity prediction. Given the length of the paper I thought I would share my important takeaways.…
May 13, 2025

Chebyshev Scalarization Explained

I’ve been reading about multi-objective optimization recently.1 Many papers state limitations of “linear scalarization” approaches, mainly that it might not be able to…
May 12, 2025

Reminder that Claude is really good for Chinese grammar

This shouldn’t surprise anybody who uses LLMs a lot, but LLMs are really good at common languages and translation. When practising Chinese today, a sentence which confused…
May 8, 2025

Some Beijing travel tips

Earlier this year I went to Beijing. Here are some miscellaneous travel tips. What I am writing here is probably not unique or original, treat this just as my personal…
Apr 28, 2025

Taking the V out of VAEs: long live KL-regularized autoencoders!

I recently read this post about generative modelling in latent space by Sander Dieleman and agreed strongly with the following quote (copied verbatim except for typo…
Apr 24, 2025

Coding python packages with AI

I tried using some new LLM tools to code 2 entire python packages (instead of editing a handful of lines at a time, which is what I did previously). It went well! These…
Apr 12, 2025

Why your active learning algorithm may not do better than random

I am a big fan of active learning, but I am also acutely aware of its potential failure modes. A common failure mode is random-like performance: achieving no better “success”1…
Apr 2, 2025

Generic recommendations for cheminformatics models

I was recently asked via email for my thoughts on which models to use in general for molecular property prediction. I’m sharing my responses publicly in case they are useful.
Apr 1, 2025

Using LLMs to improve my Chinese

I’ve been learning Chinese for almost 10 years now, but still make awkward-sounding sentences when I speak. A few months ago I thought “why not use LLMs to help me speak…
Mar 30, 2025

Conceptual confusion about desirable outputs of reaction prediction models.

In the literature about machine learning for retrosynthesis, one line of work tries to predict chemical reactions, either in the forward direction (ie what products will A +…
Mar 27, 2025

Punishing poor reviewers at CVPR

This year CVPR pledged to make all authors participate in peer review, and reject papers from authors who wrote low-quality reviews.1 Last week they confirmed on Twitter that…
Mar 4, 2025

Why don’t ML conferences provide reviewer instructions?

I remember when I first received an invitation to review papers for an ML conference in late 2020. What surprised me most was not that I was being invited (even though that…
Feb 25, 2025

Alpha over LLMs

On a recent podcast, Patrick McKenzie mentioned the idea of “alpha over LLMs”: does a publisher produce text with any meaningful advantage over asking an LLM? I think this…
Feb 24, 2025

Is offline model-based optimization a realistic problem? (I’m not convinced)

This is a “quickpost”: a post which I have tried to write quickly, without very much editing/polishing. For more details on quickposts, see this blog post.
Feb 14, 2025

Experiment: more posts, lower quality

Since starting my new position at Valence, my efforts to write more on my blog have clearly been successful.1 However, my internal list of “things I would like to write a…
Feb 11, 2025

Stock responses about statistical significance for reviewing machine learning papers

So many ML papers contain tables like
Feb 11, 2025

Hiring is hard: why good applicants without connections can get overlooked

Knowing people is a great way to get hired. Nepotism is one obvious explanation (aka people hire you because they like you, or to gain favors from people who like you). I…
Feb 9, 2025

Reaction model scores are CRITICAL to multi-step retrosynthesis

Machine-learning for retrosynthesis is a popular research topic. Popular sub-topics include:
Jan 26, 2025

Double checking that Gauche’s fingerprint kernels are positive definite

GAUCHE is a library for Gaussian processes in chemistry. I contributed a small amount to GAUCHE several years ago but am not an active developer. I recently learned that…
Jan 12, 2025

What ML researchers and users get wrong: optimistic assumptions

ML is often done poorly, both by “ML experts” (by which I mean people who understand the algorithms but not the data) and “ML users” (by which I mean people who understand…
Jan 9, 2025

New Year’s Resolutions for 2025

Happy 2025! Here are a few goals I am setting for myself this year!
Jan 5, 2025

Review of NeurIPS 2024 and predictions for ML in 2025

I was fortunate to attend NeurIPS 2024, arguably the largest and most influential machine learning conference in the world (thanks Valence for sponsoring my trip 🙏). In…
Jan 1, 2025

Rules of scientific English writing for an international audience.

Although English is the common language for international scientific communication, most scientists are not native English speakers. To account for this, I think that all…
Dec 30, 2024

When should you expect Bayesian optimization to work well?

As much as I believe in the potential of Bayesian optimization (BO) to be useful for scientific discovery, after 4+ years I have seen many instances where BO does not work.…
Dec 10, 2024

What can eduroam teach us about building research infrastructure

Eduroam is a fantastic piece of academic infrastructure: students/researchers from thousands of universities around the world can automatically connect to WiFi and any…
Dec 1, 2024

Scientific conferences as approximate Bayesian inference

Scientists should ideally form their beliefs based on evidence and update their beliefs as new evidence arrives. Unfortunately, humans are far from perfect Bayesian thinkers…
Nov 17, 2024

Overview of the NSGA-II algorithm for multi-objective optimization

Nondominated sorting genetic algorithm version 2, more commonly called NSGA-II, is a well-established genetic algorithm (aka evolutionary algorithm) for multi-objective…
Nov 4, 2024

Advice for applying for a PhD in AI for science in 2024

With the immense growth of AI for science (AI4sci for brevity), I imagine many younger students are considering applying for PhDs in AI for science in this application…
Oct 27, 2024

Thoughts on Google Vizier

Vizier, described in a recent paper from Google, is a black-box optimization algorithm deployed for “numerous research and production systems at Google”. Allegedly, this one…
Oct 13, 2024

Being ‘data-driven’ does not mean that you should use bad data.

Relying on data rather than intuitions to make decisions is usually a good thing, but is not always better. When one needs to make a decision about things for which there is…
Oct 9, 2024

Problems with the top-k diversity metric for diverse optimization

NOTE this blog post can be run as a jupyter notebook. I re-ordered the cells to make it easier to read; to re-produce all the plots see instructions at the end of the post. Background ’Diverse optimiz…

Oct 7, 2024

How I chose a static site generator

Recently, I wanted to update my website to look a bit more polished (and support additional features such as automatically generating pages for my publications). In the end…
Sep 9, 2024

Reasons to have a website

I created this website because I thought (and continue to think) that having a website can benefit one’s career. Essentially, a professional website serves as an accessible…
Aug 26, 2024

The Monty Hall Problem is Really About Policy Assumptions

TL;DR: the Monty Hall problem doesn’t have a well-defined solution.
Feb 10, 2020

A Quick Tutorial on Bash Quotes

Today I learned way more about quotations in bash than I ever thought I needed to know. I thought I would highlight the interesting use case that I discovered, which…
Jul 18, 2019

How to Keep a Communal Fridge Clean

Last month, my class decided that we should get a fridge for the class study room. This brought up an important question: how would the fridge be cleaned? I thought this was…
Oct 8, 2018

Language Travel Logs: Japanese 2018

One dream I have always had since I started learning languages is to be able to go to another country and use that language to communicate. This August I had the first…
Sep 19, 2018

Sparse Matrices: 5 Tips and Tricks

Over the course of my internship at the online shopping company Wish, I have dealt a lot with a lot of data in the form of sparse matrices, specificaly in the form of item…
Sep 12, 2018

Why you should never be certain of your beliefs: a Bayesian perspective

People are notoriously bad at estimating their percent confidence in their beliefs, as explained further in this Wikipedia article. Something I thought of recently is what…
Aug 21, 2018

An Overview of Gradient Boosting and Popular Libraries for it.

Everybody doing machine learning wants the best models possible. The aim of this blog article is the following:
Jul 17, 2018

Turning Adam Optimization into SGD

This strange question came up when working on a machine learning project to generate embeddings. Working with the version of Pytorch available on our DGX (similar to version…
Jul 2, 2018

Paper Review: A Computational Approach to Organizational Structure

Ever since I’ve started doing internships, the concept of efficient organizations has piqued my interest. In every workplace I have been in, time is always wasted by…
Jul 1, 2018

How to guess a Kanji’s on-yomi in 4 easy steps

Lately I have been putting a lot of effort into studying Japanese, to prepare for my upcoming trip to Japan at the end of summer 2018. While learning the readings for…
Jun 25, 2018

Launched my website

Didn’t know that creating a decent looking website could be this easy. Thanks Jekyll. Also thanks to this website for walking me through the process.
Mar 24, 2018
No matching items