Yi–Ning Weng

Vision

I work toward a future where human centered intelligence becomes a force that lifts the world forward. In this future, technology deepens our capacity to understand one another, enriches the cultural opulence of our societies, and turns creation itself into a path of collective growth. I believe abundant intelligence can become a gentle presence that expands patience, wisdom, and possibility, guiding real improvement for humanity.

Areas of Focus

Reinforcement Learning ・ Multi-agent Systems ・ Complex Systems

Publications

Research

I study how agents can develop long-term strategiesthat maximize collective rewards rather than short-term gains in dynamic games, modeling their behavior with Q-learning and Deep Q-Networks on evolving network topologies.

I have introduced several approaches to strengthen agent adaptivity. I explored more granular action spaces that allow agents to take neighbor-specific actions, enabling finer control of local dynamics. I also developed reinforcement-learning-based rewiring rules, through which agents can reshape their neighborhoods by forming or severing links based on long-term strategic planning. These methods help stabilize cooperation, increase equilibrium cooperation rates, and improve learning stability in large-scale grid and power-law networks.

During my undergraduate studies, I conducted research at Academia Sinica in collaboration with Prof. Hsuan-Wei Lee, who has held appointments at Academia Sinica and Lehigh University and previously taught at National Taiwan University. We have completed two papers, with another currently in progress.

More details ↗

Projects

★

Auralytics: Your fascinating music stats for Spotify

Auralytics is an open-source project designed to provide in-depth insights into users' Spotify listening habits, supporting 10 languages. By leveraging Spotify's API, Auralytics provides users with their top tracks, artists, albums, genres, and musical eras.

ReactNode.jsRedis

👥 1

Machine Learning Translation for the Puyuma language

This project fine-tunes Hugging Face’s mT5 model in Python/PyTorch to translate between Chinese and Puyuma, an indigenous low-resource language with only 2k sentence pairs. It applies customized preprocessing and lexicon-based augmentation to enhance translation accuracy.

mT5PyTorchTransformer

👥 2

Predicting Baseball Season Results with Machine Learning

This project builds machine learning models in Python to predict baseball season outcomes. Through feature engineering, hyperparameter tuning, and model optimization, it achieved top 13% performance (18/141 teams) in a Kaggle competition.

scikit-learnSVMRandom Forest

👥 2

yiningweng.com

A personal website built with Gatsby (React + GraphQL) to share original articles, photography, and podcasts. Leveraging static site generation and automated deployment, it delivers a fast, scalable, and engaging browsing experience for over 2,000 unique users.

GatsbyReactJavaScript

👥 1

The Adventurous Journey of Business Students in the Future World

an interactive visual novel created with Ren’Py to explore future challenges and opportunities in business education. By combining storytelling and decision-based gameplay, the project encourages students to develop critical thinking, innovation, and teamwork skills through immersive scenarios.

C++

👥 4

SmartSpace: An innovative furniture solution optimizing limited living space

an innovative startup project that designs smart furniture solutions for compact living. By integrating design thinking and business modeling, it reimagines urban living through space-efficient furniture and secured partnerships with leading construction companies.

startup

👥 7

Professional Journey

From business school to engineering and machine learning, my path has always been defined by intention and discipline.

Beyond Engineering