Neural Playground | AI Visualizer

01 / CONFIGURATION

Network Setup

Customize your neural network architecture.

Input Neurons

Hidden Layers Enter neuron counts separated by commas.

Output Neurons

Weight & Bias Initialization

Initialization mode

Random seed Use the same seed to reproduce an experiment.

02 / SIMULATION

Model Settings

Activation Function Hidden layers use this function. The output layer uses Softmax.

Output Layer Activation Softmax operates across the complete output vector and produces probabilities that add up to 1.

Loss Function Per-output mean: average squared difference between target and prediction.

Softmax + Cross-Entropy L = −Σ yᵢ log(pᵢ) Softmax converts output logits into probabilities that add to 1. Cross-entropy measures how much the predicted distribution differs from the target distribution.

Input Values Example: 0.8, 0.2

Target Values Expected output probabilities; values must add up to 1.

Learning Rate

Train multiple iterations

LIVE VISUALIZATION

Network Architecture

LIVE

Drag to rotate · Wheel to zoom · Click a neuron to inspect

Input Layer Hidden Layer Output Layer Active signal

LIVE EXECUTION

Calculation Panel

MATH ENGINE

Waiting to start simulation...

READY Awaiting a simulation

CURRENT STEP

Ready to explore

Configure your network and run a simulation. Manual Mode pauses after every teaching step.

Step history 0 steps

Previous steps will appear here.

INPUT DATA

Input Values

Waiting for input...

MODEL OUTPUT

Prediction

Waiting for simulation...

ITERATION SUMMARY

Before vs After Training

Run an iteration to compare the updated prediction and loss.

PARAMETERS

Current Weights & Biases

Generate a network to inspect parameters.