I'm a machine learning researcher specializing in deep learning architecture design. I use this site to show some of the previous projects I've worked on.
Quick post about a diagram I released on X a few weeks ago! It's DeepSeek-V3 - a. On the leaderboards , it's performing better than the May version of GPT-4o, indicating that open-source models are less than 7 months behind OpenAI. Here's the
This post covers the foundations of diagrams—how they represent data types and the functions between them—with minimal fluff. There is no category theory in this post, which may encourage or dissuade you depending on your background. If details are missing, I've oversimplified; if too
Many papers in the field are rather indecipherable, not providing in prose, diagrams, or equations all the information needed to know what an algorithm actually does. I believe this comes down to the difficulty of expressing the exact size of data throughout a model and how operations are broadcast to
This is drawn from Section 1.3 of my paper Neural Circuit Diagrams: Robust Diagrams for the Communication, Implementation, and Analysis of Deep Learning Architectures . It is a primer for a series of posts I am about to make on how Neural Circuit Diagrams can address many pressing problems in
I have recently been working on papers for the Applied Category Theory Conference 2024 with Prof. Gioele Zardini . Here, I've got a link to the two papers I have been working on. The first, Functor String Diagrams , covers the mathematics behind the streamlined approach to diagrammatic category theory
This blog post is adapted from an X thread I posted. Its garnered significant interest, so I decided to post it here as well! Mixtral-8x7b by @MistralAI is an LLM that outperforms all but OpenAI and Anthropic's most powerful models. And, it is open-source . In this
Earlier this year, I had to take a research skills course. I used this opportunity to focus on two things in deep learning that really stood out to me. Firstly, skip connections completely change the way models operate. They make deep learning possible by changing the task of models into
This is a slightly edited version of a report I wrote for a summer research internship I did in 2021/2022 (remember, seasons are swapped in Australia). It captures why I'm skeptical of specific mathematical explanations for specific behaviors of deep learning models. It is why I am
This post does not directly deal with deep learning – rather, it's a passion project from last year when I was obsessed with understanding the physics of entropy better. Entropy is a concept from thermodynamics that snuck itself into computer science following the work of Shannon. The second
Robust Diagrams for Deep Learning Architectures Robust Diagrams for Deep Learning Architectures.pdf 7 MB download-circle I have recently finished my honors thesis. An honors thesis is something of a mix between a senior
As some of you might know, I have been working on neural circuit diagrams over the past year or so. These diagrams solve a lingering challenge in deep learning research – clearly and accurately communicating models. Neural circuit diagrams are a mathematically robust framework to precisely express how components interact
I'm Vincent, I spend my time learning maths and finding new ways to apply it. For the past few years, I have been looking at deep learning models and whether better ways exist to understand them. Deep learning is a field in far more flux than people realise,