Introduction
Hi there I’m dzungphieuluuky. I’m currently an undergraduate student and I’m seeking for opportunity to do research in deep learning and intelligent systems. This is my portfolio website to update my achievements and share my experience and thoughts through blogging.
I’m fascinated by how physics intuition can help us build intelligent systems that are arising in deep learning and the emergence of a new field of learning mechanics.
Research Interests
Currently, my main interests are:
- Deep Reinforcement Learning: Sample efficiency, exploration-exploitation trade-offs, dream models that can learn in their own dreams :D
- Representation Learning: Designing intelligent regularization techniques for complex models, innovating information fusion methodologies across different modalities to boost performance and versatility.
- Diffusion Generative Modeling: Diffusion-based generative models, implicit regularization in their training dynamics, and how can we use them to bridge the simulation-to-real gap in robotics learning.
Favourite Channels
Watching YouTube is one of my ways of burning time when I’m free, some of my favourite channels are:
- 3Blue1Brown: a must-watch channel for those who love mathematics with strong intuition. I’m especially fascinated by his series on Essence of Calculus and Linear Algebra. Mathematics in the hand of Sanderson is truly something godlike and out-of-this-world.
- Veritasium: a channel for anyone burned continuously with curiosity and desire to expand knowledge.
- Fireship: coding channel with lightspeed way of explaining things, suitable for those who needs some speed to focus.
Favourite blog posts
- Sander Dieleman: a research scientist at Google Deepmind. I’m currently learning a lot from his blog posts, which range around 30 minutes to 60 minutes reading time. This is a treasure trove for those who seeks to understand generative modelling far beyond the textbooks as the insights from his research is totally fantastic.
- Lilian Weng: I usually use her blog posts to graps the broad understanding of the state-of-the-art of any machine learning disciplines that I’m interested. Her blog posts are very long, containing a large amount of knowledge about many topics. If you want to grasp what is happening in a specific field, her blog posts are definitely where to go.
- Cameron Wolfe: This LLM researcher has a nice substack portal with dozens of high quality blogs suitable for those interested in deep learning frontiers in general and LLM in particular. I find his posts contain a handful of knowledge without being too mathematically heavy to understand. Good starting point for those who are beginners in the field or not ready for more math-heavy blogs.
Mathematics taste
- Like: linear algebra, calculus, differential equations and high-dimensional unintuitive phenomenona.
- Don’t like: combinatorics, graph theory, number theory, discrete mathematics in general.
What have I been up to
Font Architect - Diffusion Models
A research project where I try to develop methods for generating high quality images for Sino Nom language, an ancient language in historical sripts. The generated image blends the content of one input image and style of another input image. Long story short, the model needs to learn how to fuse the content features from an image with the style features from another to generate an image that has the content of image A but written in styles of image B. Some major difficulties happen along the way when the model tries to capture the style features from the content image and blends it with the style features of the other and create the final, messy image (which is not what I desired). This project is the one that has sparked my interest with diffusion models recently since I have no prior experience or knowledge about diffusion generative modelling before exploring this projects. Because of this project, diffusion models has just been appended to my interest list which makes it longer than ever.
Ouro Trace - Looped Language Models
This is my capstone project for the course Introduction to Machine Learning. In this project, I experimented with a small looped language model, ByteDance/Ouro-1.4B-Thinking · Hugging Face, which is one member in the Ouroboros family developed by ByteDance. I managed to survive through a lot of difficulties doing this project due to my inexperience with natural language processing. All of my previous projects are mainly around reinforcement learning so this project was some kind of a challenge :D. I think the difficulties also come from the different nature of looped transformer architecture compared to traditional transformers. Learnt a lot, definitely.
Energy Management - Deep Reinforcement Learning
I have done this project as part of my participation of Viettel AI Race competition, where I had to develop an inteligent agent to autonomously control and orchestrate the output burst of cells in a energy grid depending on the energy demands at hand. This challenge is quite hard since I was doing it alone along with several courses (I registered 6 courses in the same semester beside this thing, quite a lot of work to do). Learnt quite much about the behaviour of PPO and SAC training, experience the notoriously famous instability of RL training with my own eyes :D
This website
A beautiful template from Beautiful Jekyll where I’m going to share some of my personal experience while I was doing my projects and some interesting thoughts (I think) about developing intelligent architectures and loss functions.
Fun facts
- I’m most productive during night time.
- I enjoy reading anime character wikipedias rather than actually watching that exactly anime series. Quite excited to do research about my favourite anime characters.
- My learning fuel: curiosity and probably too much coffee everyday intake.
- I’m apparently a generalist rather than a specialist (maybe).