Computation Graph

Computation Graph — Micrograd

Interactive DAG for loss = (w·x − y)². Drag the w slider; watch values and gradients update live. Enable Animate Backprop to see the gradient flow direction.

Weight w0.5

Show Gradients

Animate Backprop

Loss = 4.0000

Optimal w = 1.5 (y/x = 3/2)

One GD step (lr=0.1):

w → w − lr·∇ = 0.500 − 0.1×(-8.000) = 1.300

Forward pass

Backward pass (gradients)

val= forward value

∇= ∂loss/∂node

Hover any node for details. Drag w toward 1.5 to minimise loss. At w = 1.5, w·x = 1.5·2 = 3 = y, so loss = 0.