Neural Network Lab

Feedforward MLP · tanh hidden · sigmoid output · MSE loss

Signal propagation

Animation speed 1.0×

Input sample

Training · backpropagation

Target class

Learning rate η 0.60

Steps

Prediction

–

Sample loss

–

Dataset loss · acc

–

Architecture

Hidden layers

Neurons per hidden layer 6

Legend

−w0+w

Negative Positive Gradient δ Changed by edit

Drag to orbit · scroll to zoom · right-drag to pan.
Hover a neuron for its state. Click a synapse or neuron to edit it.
Space flow · T train step · Esc close