MLP Playground

Train a configurable multi-layer perceptron on clustered 2D points

Architecture

Hidden Layers
Layer 1
4 neurons
Layer 2
4 neurons

Training

Metrics

Epoch
0
Loss
Accuracy
Classes
3

Data Distribution & Predictions

Points are sampled from distinct clusters in 2D space. Fill color shows the predicted class, while the outline denotes the actual class. Larger outlined points mark misclassifications.

Class 0
Class 1
Class 2

Network Topology

Edge thickness encodes weight magnitude; color shows sign. Hover over nodes and connections to inspect exact values. Biases appear inside nodes, and small networks surface numeric labels directly on each edge.

Hover a node or edge to reveal exact weights, biases, and connectivity counts.

Weight Snapshot

Initialize the network to inspect its weights.

0.0 ms