Interactive 2D simulation · Artificial intelligence
Neural Network Gradient Descent Studio
Train a small neural network to fit a noisy nonlinear dataset and watch gradient descent reduce prediction error.
LIVE MODEL
LIVE EXPERIMENT VIEWCANVAS RENDER · REAL-TIME MODEL
What you are exploring
A neural network learns by adjusting weights to minimise a loss function. This one-hidden-layer model uses tanh activation and gradient descent on a synthetic regression dataset. Learning rate controls the size of each parameter update; too small is slow, while too large can overshoot.