# Forward Pass

Canonical: https://manyspace.aporia.systems/tools/neural-network
Updated: 2026-10-13

Two inputs, three hidden neurons, one output — nothing hidden. Push an input and watch every neuron answer at once.

## Try this

- Wake the silent one. Find a hidden neuron reading “0, off”, then sweep x₁ until it fires — ReLU neurons sleep below zero and wake above it.
- Kill the biases. Toggle biases off and watch every z shift by exactly the bias amount — small numbers steering big decisions.
- Swap the squash. Switch to sigmoid: every neuron now answers in shades of gray, and “firing” stops being all-or-nothing.

## How to read the network

Each hidden neuron forms a weighted sum of the inputs, adds its bias, and squashes the result: ReLU keeps positives and silences negatives, sigmoid bends everything into 0–1. Edge thickness carries the weight, node faces carry the live values.

The output neuron does the same trick once more over the hidden activations. Flip the activation or kill the biases and re-sweep: the same inputs now tell a different story, and you can see exactly where it changes.

## Related tools

- [Gradient Descent Playground](https://manyspace.aporia.systems/tools/gradient-descent)
- [Loss Function Explorer](https://manyspace.aporia.systems/tools/loss-functions)
- [k-Means Visualizer](https://manyspace.aporia.systems/tools/k-means)
- [Logistic Regression](https://manyspace.aporia.systems/tools/logistic-regression)
- [Backpropagation](https://manyspace.aporia.systems/tools/backpropagation)
- [Attention Visualizer](https://manyspace.aporia.systems/tools/attention)
- [Tokenization Explorer](https://manyspace.aporia.systems/tools/tokenization)