Neural Network Playground

Explore a real feedforward neural network interactively — adjust inputs, weights, bias and activation function and see the classification decision boundary recompute live.

A single artificial neuron computes a weighted sum of its inputs plus a bias (z = w₁x₁ + w₂x₂ + b), then applies an activation function — the logistic sigmoid σ(z) = 1/(1 + e^−z), which squashes any real number into (0, 1), or the rectified linear unit ReLU(z) = max(0, z) — to produce its output. Comparing that output to a classification threshold (usually 0.5 for sigmoid) turns it into a binary decision. The default network here is `@scienceverse/simulations-core`’s own real `AND_GATE_NETWORK`: with large weights (20, 20) and a negative bias (−30), the sigmoid saturates toward its 0/1 extremes and the neuron approximates the logical AND gate — a standard introductory machine-learning construction (Nielsen, M., Neural Networks and Deep Learning, ch. 1; Andrew Ng’s Machine Learning course material on logic gates as perceptrons), computed by that package’s real, tested `neural-network` model, not an approximation invented for this page.

This is a forward-pass-only playground — there is no training or backpropagation here (PRD §12.1’s own scope for this simulation: “inputs, weights, activations, classification boundary”). It is intentionally a different thing from the separate Neural Network Explorer 3D scene: that scene visualises a fixed multi-layer forward pass in 3D, while this simulation lets you directly manipulate the weights, bias, activation function and classification threshold yourself and watch the classification decision boundary — the shaded region of the input plane the network calls “true” versus “false” — change live, sampled across a real grid of inputs in a dedicated Web Worker. A shareable link captures the exact current network. Readers whose browser cannot run canvas rendering, who prefer reduced motion, or who have JavaScript disabled see the real computed boundary and layer outputs below instead, in an accessible diagram and data table.

This interactive simulation is part of ScienceVerse Plus.

Everything you need to understand the idea is on this page. Plus and Family plans add the hands-on explorer, along with higher tutor limits, full guided learning paths and progress exports.