Timothy Lee's repositories
R-Projects-in-Time-Series-Analysis
A collection of time series analysis and modelling projects using R. Models implemented includes (seasonal) ARIMA models, multivariate VAR models, GARCH, and ARMA-error regression model with other external regressors.
R-Projects-in-Data-Analysis
A collection of data analysis and modelling projects using R. Models implemented included GAMs, GAMMs, GLMs, GLMMs, LMMs, MLRs, and case control studies with both cross-validation and maximum likelihood estimation, accounting for random effects for individual groups if needed.
Android-Game-Center-App
An android-based game app designed and implemented with a team of 4. Includes a sliding tiles game, 2048, and sudoku. Various OOP principles and design patterns were implemented alongside with intuitive UI/UX design.
K-Means-GMM-and-Reinforcement-Learning
Implemeted a Q-learning algorithm to find a shortest path in a stochastic maze under a Reinforcement Learning context. Experimented with k-Means and EM algorithm to find data clusters.
TrueSkill-Ranking-Model-Using-MCMC-On-Tennis-Players
This project demonstrates Stochastic Variational Inference in the TrueSkill Model, using Bayesian inference in large models with continuous latent variables and Markov Chain Monte Carlo on tennis data sets.
AI-Search-Algo-Sokoban-Game
An AI search algorithm to solve the Sokoban Puzzle involving multiple agents (robots).
AI-strategy-games-Stonehenge-and-Subtract-Squares
2 AI-based strategy games
AI-to-Player-OthelloGame
An AI-to-Player Othello Game with GUI on different board sizes. Uses both minimax and alpha-beta implementation with depth limit, state caching, and node ordering heuristics.
BayesNet-and-Bayesian-Inference
Implemented variable elimination for any given Bayes Net. Used the net for problem-solving and inference.
covid-19-excess-deaths-in-Latin-America
Inspired from The Economist article on “Tracking covid-19 excess deaths across countries” (The Economist, 2021), this report aims to investigate the proportion of overall excess deaths vs the official reported death counts due to the covid-19 pandemic using a Bayesian approach.
Decision-Trees-Identifying-Fake-News-Headlines
Using the scikit-learn decision tree classifier to classify real vs. fake news headlines
Julia-Intro
Brief intro to using Julia as a numerical computation language, including experimenting with automatic differentiation (auto-diff).
Julia-Neural-Network-with-Nonlinear-Regression-and-heteroscedastic-noise
This project experiments with fitting linear and non-linear regression models with heteroscedastic noise using multi-layer neural networks and MLE to capture both the mean and variance.
K-Fold-CV-Experimentation-in-Ridge-Regression
The goal of this mini project is to experiment with the K-fold cross validation (CV) procedure to tune the penalty parameter λ in Ridge regression, without using any pre-defined packages from libraries like scikit-learn.
Logistic-Regression-and-KNN-on-MNIST-digits
Implemented and optimized both k-NN and penalized logistic regression model to classify hand-written digits of '4's' and '9's'. Plotted visualizations for both models in terms of cross entropy loss and classification error with different hyper-parameters for comparison between models.
ML-Logistic-Regression_K-NN
Classify hand-written digits of '4's' and '9's' using KNN and logistic regression
NN-Identify-Facial-Expressions
Used training data obtained from Toronto Faces Dataset to train a multi-layer Neural Network for various facial expression detection via ReLu and backpropagation with weight updates and momentum
PCA-and-Naive-Bayes-Classifier
Projected training data of ’2’s’ and ’3’s’ to lower dimensions and used a 1-NN classifier on the first K principal components. Experimented the same classification with Naives Bayes Classifier
Variational-Autoencoder-on-MNIST
This project implements and investigates the Variational Autoencoder on binarized MNIST digits by building a generative model to infer the bottom half of the given binarized MNIST digits conditioned on the top half of these images.
Inforum-content-marketplace
Inforum is a promotional content marketplace that aims to market to hard-to-reach millenials/Gen-Z demographics by utilizing network effects on social media platforms such as Facebook, Twitter, and Instagram.
covid-19-excess-deaths-tracker
Source code and data for The Economist's covid-19 excess deaths tracker
data-science-question-answer
A repo for data science related questions and answers
light-bootstrap-dashboard-react
React version of Light Bootstrap Dashboard
Neur2SP
Neur2SP: Neural Two-Stage Stochastic Programming
sktime
A unified toolbox for machine learning with time series