Home▸Machine Learning & Neural Networks▸Autoencoder Anomaly Detection: Reconstruction Error Simulator

🧠 Autoencoder Anomaly Detection: Reconstruction Error Simulator

Watch data stream through an encoder-decoder neural network, get compressed through a bottleneck, and get judged by reconstruction error. Tune the bottleneck size and threshold to see how autoencoders flag anomalies.

Machine Learning & Neural Networks3DAdvanced60 FPS

Details

CategoryMachine Learning & Neural Networks

RenderingThree.js · WebGL

DifficultyAdvanced

Target FPS60

ControlsBottleneck-size slider (2-16) changes the compression width and the reconstruction-quality/anomaly-detection trade-off; anomaly-rate slider (0-40%) sets how often a genuinely anomalous point enters the stream; threshold-percentile slider (80-99.9) sets the statistical mean+z*sigma cutoff used to flag anomalies; stream-speed slider changes how fast data points are generated and pushed through the network; Pause/Resume button freezes the stream.

Three.jsautoencoderanomaly detectionneural networkreconstruction errorInstancedMesh

📖 Read the articleThe maths behind Autoencoder Anomaly Detection: Reconstruction Error Simulator

Related simulations

Decision Tree LiveCART decision tree with Gini impurity splitting; axis-aligned splits emerge level by level…

Reinforcement LearningQ-learning agent navigates an 8×8 grid maze. Heatmap shows max Q-values; white arrows show…

Self-Organising Map24×24 Kohonen SOM (576 neurons, 3D RGB input). Gaussian neighbourhood update as α and σ… ⊞ All 165 in Machine Learning & Neural Networks →

⚙ Under the hood

Stream data through an encoder-decoder neural network, watch it get squeezed through a bottleneck, and see reconstruction error flag anomalies in real time. Tune bottleneck size, anomaly rate and the statistical threshold to see the detection trade-offs play out.

Three.jsautoencoderanomaly detectionneural networkreconstruction errorInstancedMesh

3D · Three.js / WebGL renderer · 60 FPS target · runs fully client-side, no install