Afterglowscreensavers

Neural Network Forward Pass Screensaver

The Neural Network Forward Pass screensaver runs a real small feedforward neural network live, propagating random inputs through its actual weighted layers. It is free, runs in your browser, and ESC exits.

Live preview

Press ESC to leave fullscreen. Settings are saved in your browser only.

How the Neural Network Forward Pass screensaver works

The saver builds a real small feedforward network with a fixed random set of weights across a few layers of neurons.

A real random input vector feeds into the first layer, and every neuron's activation is genuinely computed as a weighted sum passed through a real activation function.

Each layer's real activation values light up its neurons, and the brightness of each connecting line reflects its real weight and signal strength.

Once the forward pass reaches the output layer, it holds briefly to show the result, then a fresh random input begins propagating.

A worked example

A strongly positive input paired with a strongly positive weight lights a connection brightly, while a negative weight on the same input visibly dims or suppresses the neuron it feeds — real weighted computation, not decoration.

Settings & tips

Frequently asked questions

Is this a real neural network?
Yes — a real small feedforward network with genuine weights and a real activation function, not a stylized animation.
Does it learn or train?
No — this shows a real forward pass with fixed weights; it does not adjust weights through training.
Is a key needed?
No — it is a pure local computation.
Is it free?
Yes — free, no download, in your browser.

← All Science & Research screensavers