Icon

Autoencoder for anomaly detection

Step 6. Define classes
Step 2. Design the Keras model (Autoencoder 140-100-70-100-140)
Step 6. Optimize & evaluate model accuracy

Anomaly Detection with Autoencoder


Step 1. Data preparation & partitioning
Step 3. Train the neural network
Step 4. Apply trained network
Step 5. Calculate MSE of reconstructed features


Apply trained network
Keras Network Executor
Units: 70Activation: ReLU
Keras Dense Layer
Normalizes and partitions training, validation and test data
Preprocessing
Evaluate model
Comparison
Keras Network Writer
min_threshold
Expression
Train with Loss function=MSE Optimizer=Adam
Keras Network Learner
Number Rounder
Table Row to Variable
max_threshold
Expression
Savenormalizationmodel
Model Writer
Calculate MSE
Java Snippet
Units: 140Activation: Sigmoid
Keras Dense Layer
Reconstruction errors distributions
Shape: 140
Keras Input Layer
Units: 100Activation: ReLU
Keras Dense Layer
Only normal heartbeats
Row Filter
Units: 100Activation: ReLU
Keras Dense Layer
Table Reader
Denormalizes true and predicted features of an a-normal heartbeat and visualizes them in comparison
True vs Predicted
Heartbeats

Nodes

Extensions

Links