Loss landscape explorer

Change learning rate and momentum. Follow actual gradient-descent steps across an interactive 3D surface.

How this experiment works

Analytic derivatives update two parameters through gradient descent with optional momentum. The complete loss trace is measured.

What it does not prove

This is an analytic teaching function with two variables, not a measured neural network loss landscape.

Try this challenge

Increase momentum and learning rate together. When does the path overshoot?

Related Trust Agent courses · Sources and licences

Enable JavaScript to run the experiment on your device.