Lê Chí ĐạiGraphics · Interactive learning · 2026
gradphic
A graphics coursework project that turns optimization into a surface, a moving marker, and a trajectory you can inspect.
Making the process visible
The application renders sampled loss surfaces as triangle meshes and displays optimizer trajectories, with interactive camera controls and an ImGui interface.
Graphics and algorithms
The rendering work includes surface geometry, lighting, trajectory lines, and animated optimizer markers. The optimizer code includes gradient descent, momentum, and Adam.
An educational visualization
This is not an optimizer benchmark. The SGD implementation uses a full gradient, mini-batch behavior is simulated with noise, and the Nesterov update simplifies the look-ahead step.
Next project