2D Neural Network Forward Pass 2

This simulator draws a fully-connected feed-forward neural network as a flat, editable diagram instead of a rotating 3D scene. Set the number of input and output neurons, choose one to three hidden layers with their own neuron counts, and pick an activation function — sigmoid, tanh, ReLU, or linear identity. Dragging an input slider re-computes the real weighted-sum-plus-bias math for every neuron and instantly updates each node's color and brightness to match its activation, while the "Run forward pass" button animates the same computation as pulses traveling along the weighted edges from the input layer to the output layer, one hop at a time. Hovering any neuron reveals its exact pre-activation sum and post-activation value, and the output panel tracks the network's final numbers live as you experiment with architecture, weights, and inputs.