1. Configure
Each column selects a key hyperparameter. The project type determines the output shape, layers control depth, and neurons set the width of each layer.
// Interactive Playground
Neural Network Builder
Spin the columns to configure your architecture, then watch a neural network materialize, train, and converge — all in your browser.
Configure Your Network
Scroll each column to pick your parameters, or hit Randomize to roll the dice.
Project Type
Image Classifier
Text Generator
Speech Recognizer
Sentiment Analyzer
Anomaly Detector
Recommendation Engine
Object Detector
Style Transfer
Hidden Layers
1 Layer
2 Layers
3 Layers
4 Layers
5 Layers
6 Layers
7 Layers
8 Layers
9 Layers
10 Layers
Neurons per Layer
8 Neurons
16 Neurons
32 Neurons
64 Neurons
128 Neurons
256 Neurons
512 Neurons
Image Classifier
3 hidden layers
64 neurons/layer
ReLU
LR: 0.01
Network Architecture Ready
Training Dashboard Idle
--
Accuracy
--
Loss
0
Epoch
/ 0
--
Parameters
trainable
Training log will appear here...
What's Happening Here?
2. Build
We construct a feed-forward neural network from your choices. The visualization shows every neuron and the weighted connections between layers.
3. Forward Pass
Data flows left to right — inputs are multiplied by weights, summed, and passed through the activation function at each neuron.
4. Backpropagation
The error flows backward, updating each weight to reduce the loss. This is the engine of deep learning — and it runs automatically here.