Experiment 01
2D Regression with MSE Residuals
Configure the architecture, drag sliders to tune weights manually, or hit Train to run gradient descent live. The orange lines are residuals — their squared average is the MSE.
MSE loss
—
Epoch
0
Architecture
—
lr = 0.03 · full-batch GD · tanh hidden · linear out
Data (noisy sin)
Network prediction
Residuals (↕ error)
Architecture
Hidden layers: 2
Neurons per layer
Weights & Biases