About
I am a data scientist at American Express in Singapore. I am working on AI/ML methods to identify fraud risk. I hope to update more on what I do and the interesting problems I come across, or at least whatever I am allowed to discuss.
I have completed a PhD in Integrative Sciences (Statistics) at the National University of Singapore (NUS) under the supervision of Alexandre Thiery
and Matt Graham. Prior to that I did my B.Eng. (Hons) in Engineering Science at NUS specialising in Computational Science.
You can find my updated CV here.
Research interests: Markov chain Monte Carlo methods, Bayesian machine learning, inverse problems, computational finance, computational physics.
Notes
I write down what I learn, sometimes.- Interview prep: Some math brain teasers
- Interview prep: Data analysis 101
- Interview prep: Data cleaning 101
- A quick calculation of the '960' in Chess960
- Effects of chain thinning on ESS estimation
- A short note on dominating measures
- Reparametrising HMC and Bouncy HMC
- Inverse Transform Sampling
- Short review on tuning L in HMC
- Slipshod summary of several papers
- Constrained HMC formulation
Livestreaming
In the final year of my PhD (2021-2022), I sometimes stream myself working on Twitch and Youtube, where I go by the username Columnspaces. I hope for laymen to find it interesting to see what my research looks like, and for researchers to seek comfort in watching me struggle. In any case, here is me with cool-looking hair learning about Delayed-Rejection MCMC and taking notes as a new LyX user.
Works
- Manifold lifting: scaling MCMC to the vanishing noise regime
Joint work with A. H. Thiery and M. M. Graham. JRSS-B. - Reflected Hamiltonian Monte Carlo
Joint work with A. H. Thiery. NEURIPS BDL Worskhop 2021.
Talks + Workshops
Feel free to contact me for the slides / materials.2021
- Poster: Reflected Hamiltonian Monte Carlo (NeurIPS Bayesian Deep Learning Workshop)
- Talk: Reflected Hamiltonian Monte Carlo (Monte Carlo Methods)
- Invited talk: Bayesian inference and uncertainty quantification (NTU Machine Learning Society)
- Invited workshop: Introduction to MCMC workshop (NUS Hackerschool society)
- Talk: Manifold lifting (INFORMS-APS, Brisbane)
- Talk: Manifold lifting (Monte Carlo Methods, Sydney)
Tutoring
I am passionate in scientific communication and education. I have taught these undergraduate subjects.
Recreational Math
When I procrastinate but do not want to feel too bad about it, I make some visualisations.Matrix with some diagonal subtractions
The orange rhombus collapsing onto a line corresponds to eigenvalue subtraction from the diagonals.
Pandemic
I know nothing about epidemiology so I read the minimum and apply some common sense to simulate
disease spread, e.g. COVID19. I allow re-infection with some low probability.
Pink = uninfected, Red = infected, Blue = recovered, Green = vaccinated, Black (not moving) = Dead. Green blocks = Vaccination centers.
Double pendulum
A classic example of a chaotic dynamical system solved with a (numerical) symplectic
integrator to properly conserve the total energy of the system.
Neural network classification
A small neural network finding the decision boundary to a binary classification problem.
Neural network regression
A small neural network fit to noisy sine wave data.
To-simulate-list: Perlin noise drawings, boids, moire patterns, spontaneous sync etc.
Recreational Non-math
I enjoy puzzles in different shapes and forms.