Home▸Machine Learning & Neural Networks▸Backpropagation — Watch Gradients Flow Through an MLP (2D)

Backpropagation — Watch Gradients Flow Through an MLP (2D)

Canvas2D backprop lab: δ-signals pulse backward through a small MLP, gradients light up each weight, and the decision boundary morphs to fit whatever data you paint on the plane.

Machine Learning & Neural Networks2DModerate60 FPS📱 Mobile-adapted⇄ 3D version

This 2D companion runs the same backprop math as the 3D version — forward pass, chain-rule backward pass, weight updates — through a plain Canvas2D view instead of a WebGL scene: pick an activation and output mode, paint your own two-class dataset (or load a preset like XOR or spirals), then step forward/backward or let SGD run continuously while the decision boundary and glowing gradient edges update live.

⚙ Under the hood

Canvas2D backpropagation lab: δ-signals pulse right-to-left through a small MLP, edge brightness shows |∂L/∂w|, and the input-plane heatmap re-renders after every weight update.

backpropagationchain rulemlpgradient descentcanvas 2d

2D · HTML5 Canvas 2D · 60 FPS target · runs fully client-side, no install