Research

Research project that I have worked and am currently working on:

Virtual Neural Lab: Making Artificial Brain

Using Deep Reinforcement Learning to create artificial brains by imitating naturalistic agents and solving the inverse kinematics task.

Deep Reinforcement Learning Imitation Learning Generative Inverse Kinematic GPU Parallel Training CPU Distributed Training

Naturalistic Imitating With State Space Modeling

Building sequence models capable of exerting naturalistic controls for long sequences.

Imitation Learning State-Space Models Deep State-Space Models Behavior Generation GPU Parallel Training

Virtual Neural Lab: Topology-Driven Insights from Neuromechanical Agent Modeling

Researching into the high dimensional topological space & learning representation.

Computational Topology Interpretability of DRL Unsupervised Behavior Segmentation

Lightweight Agent-based Multi-collaboration Framework

Farmework for tackling language model multi-agents collborations in an adversarial environment.

Multi-agent Inference System Cooperative/Adversarial Learning

Multimodal Model Embodied Simulation

Reverse engineering from a embodied language model's perspective.

MLLM Probing LLM-ology

Other Fun Projects

Some other project regarding data science and computational neuroscience:

Learn From Your Enemy

Full end-to-end A-MARL framework for competitive and cooperative AI training.

Deep Reinforcement Learning RL Environment Adversarial MARL

Sequential Dynamics Recommendation Systems for Business Owners

Building recommendation systems using sequential data for business owners.

Recommendation Systems Data Science Workflow Deep Learning

Personal "Chat" Database For Study Data

Building a numerical / text database for my study data.

Data Science Workflow Vectorized Database & Language Model Vector Database

Robust Ensemble Learning

Ensemble learning in highly imbalanced real-world datasets.

Data Science Workflow Machine Learning

Blue Dot Project

Leveraging the power of neural network to build an Neural Gadget Controller, inspired by the Neural Turing Machine.

Topological Analysis Computational Neuroscience

Multi-addicted State Q-agent Decision Making

Studying effects of addition computationally via discrete reinforcement learning.

Discrete Reinforcement Learning Computational Neuroscience

Laplacian Eigenmap Visualization

Intuitive visualization of dimensional reduction algorithm for educational purpose.

Educational Visualization Unsupervised Learning

Latent Planning as Optimization

Learning is about how to create structured projection space and plan with it during post-training inference.

Data Efficient Sampling Offline Deep Reinforcement Learning Long Horizon Planning Problems