AI / ML VisualizerRuns in your browser · No API keyOpen Model
← All topics

Training & Optimization

Optimizer Comparison

Compare SGD, momentum, Nesterov and adaptive updates.

Gradient descent improves a parameter by taking steps toward lower error.

Why do we need this? Models need a practical way to reduce their mistakes.

Watch the parameter move on the loss curveWatch the parameter move on the loss curveLossParameter xx 2 · Loss 4 · Learning rate 0.1
Loss curveCurrent parameter / slope
STEP 1 OF 7

Find a starting error

The point is our current parameter. Its height shows the error.

Continue exploring

Gradient Descent →Vanishing & Exploding Gradients →Weight Initialization →

No account required. Local images and model files stay in your browser. Educational models demonstrate mechanisms; they are not production predictors.