Home▸AI & Machine Learning▸Autoencoder Anomaly Detection: Per-Channel Reconstruction Error

Autoencoder Anomaly Detection: Per-Channel Reconstruction Error

Watch a real tiny autoencoder train live on correlated multi-sensor data, then see per-channel reconstruction error pinpoint exactly which sensor is driving an anomaly flag, not just an aggregate score.

AI & Machine Learning3DAdvanced60 FPS📱 Mobile-adapted⇄ 2D version

Details

CategoryAI & Machine Learning

RenderingThree.js · WebGL

DifficultyAdvanced

Target FPS60

ControlsBottleneck-width slider (2-6) resizes the autoencoder's compression layer and reinitializes its weights; anomaly-injection-rate slider sets how often a corrupted reading enters the stream; corruption-magnitude slider sets how far affected channels are pushed off the normal manifold; Training toggle freezes/resumes live gradient-descent learning; Reset counters clears the detected/injected anomaly tally; drag to orbit, scroll to zoom.

autoencoderanomaly detectionreconstruction errorneural networkunsupervised learningsensor monitoring

Related simulations

Autoencoder Anomaly Detection: Reconstruction Error SimulatorStream data through an encoder-decoder neural network, watch it get squeezed through a…

Autoencoder Reconstruction Error Anomaly Detection LabA 3D lab showing how an autoencoder learns a 'normal' data manifold and flags anomalies by…

Artificial Intelligence Explained: Neural Network Learning SimulatorWatch a small feedforward neural network learn in real time. Switch between supervised,… ⊞ All 1161 in AI & Machine Learning →

Most anomaly-detection demos flag a data point with a single aggregate score. This one goes a step further: a real 8-channel autoencoder — a tiny encoder-decoder neural network trained live by gradient descent, exactly like the ones deployed in manufacturing and IT monitoring — learns the shared correlation structure of a normal multi-sensor stream, then reconstructs every incoming reading. When an anomaly is injected into one or more channels, the per-channel reconstruction error rendered as paired 3D columns immediately shows which specific sensor is driving the alert, turning a bare anomaly flag into a diagnosis. Tune the bottleneck width to see the classic compression/detectability trade-off, inject anomalies at different rates and magnitudes, and freeze training to compare a converged network against one still learning.

⚙ Under the hood

A real 8-channel autoencoder trains live by gradient descent on correlated sensor data, then reconstructs each new reading; per-channel reconstruction error rendered as paired 3D bars pinpoints exactly which sensor is driving an anomaly flag, not just an aggregate score.

autoencoderanomaly detectionreconstruction errorneural networkunsupervised learningsensor monitoring

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