Gradient Descent Visualiser
Don't just memorise the formulas. Watch the math come to life.
Gradient Descent Skier
The "Valley of Error"
In Artificial Intelligence, the goal of a model is to make as few mistakes as possible. We visualize “mistakes” as altitude. The higher you are on this 3D map, the worse the AI is performing. The bottom of the deepest valley (the Global Minimum) represents the perfect model. This tool visualizes the mathematical landscape the AI must navigate to find that perfect solution.
The Skier (The Algorithm)
The glowing sphere represents the AI’s current state. Just like a skier in the dark, the AI feels the slope of the ground under its feet (calculating the gradient). If the slope tilts down to the right, the AI steps to the right. This repetitive process of “feeling the slope and taking a step” is called Gradient Descent.
Why Learning Rate Matters
Use the slider to change the “Learning Rate.”
Too Low: The skier moves safely but agonizingly slowly. It might take forever to reach the bottom.
Too High: The skier moves fast but might overshoot the valley entirely, bouncing back and forth chaotically (exploding gradient).
Just Right: The skier descends smoothly and settles at the bottom quickly.
Try our Free Nano Banana Pro Prompt Guide
We’ve created the Nano Pro Prompt Guide to transform you from a basic AI user into a prompting expert.
