Mark Ma.
← All work

Research · Reinforcement learning

Diabetes & Reinforcement Learning

An academic exploration of reinforcement learning for glucose management using simulated data.

Academic project · Simulation

Sequential decisions, changing state

This project explored insulin-dosing decisions as a reinforcement learning problem. A simulator supplied the environment and data for studying glucose management policies.

Method

Used Advantage Actor-Critic (A2C) as part of the modeling approach, connecting a learning agent with the simulated environment.

Scope

An academic simulation project, not a clinically validated dosing system. The code provides a view into the modeling and experimental setup.

Explore the code ↗
← Back to selected workGet in touch ↗