Weiming Mai

A light-weight Python-native pedestrian traffic simulation tool based on the Link Transmission Model (LTM). Enables modeling of pedestrian movements through complex networks with visualization tools.

Evaluating crowd flow forecasting algorithms for indoor pedestrian spaces using a synthetic dataset. Features train station simulation with multiple scenarios.

Novel model for pedestrian flow prediction using crowd diffusion theory. Features graph-based representation with GNN and online learning for real-time adaptation.

Server-Client Collaborative Distillation for Federated Reinforcement Learning. Addresses objective heterogeneity in FL while protecting data privacy.

Maritime Shipping Competition solution for AAMAS 2025. Combines OR-Tools CP-SAT solver with intelligent bidding strategies for optimal fleet scheduling.

This repository contains a complete T5-based transformer system for the ACM SIGSPATIAL GIS Cup 2025 challenge on human mobility prediction. The system predicts future trajectory coordinates given past movement patterns and temporal context.

A Neural Network Development Kit for MT4 (MetaTrader 4). Implements neural network architectures in C++ for algorithmic trading and financial prediction applications.

Low-count PET image reconstruction using deep learning methods including SAGAN, MAPEM-Net, and U-Net denoising for medical imaging applications.