Examples
In the repository link you can find multiple examples of how to use the library.
The few most prominent examples are:
- Concept-Agnostic Example -
This example show how to create a simple dataset with 3 concepts and carry out a concept agnostic scenario with
CumulativeStrategy and OneCLassSVM model. Leveraged data is randomly generated to show how to create
a
ConceptDatasetusing any numpy data. - Concept-Incremental Example - This example show how to create a simple dataset with 4 concepts and carry out a concept incremental scenario with CumulativeStrategy and IsolationForest model. Leveraged data comes from different normal distributions to showcase forgetting and knowledge retention.
- Concept-Aware Example - This
example showcase how to run a concept aware scenario using the NSL-KDD dataset
stored in
resourcesdirectory. The scenario includes IsolationForestAdapter model and Replay strategy. - UNSW Dataset Example - This example showcase how to run a concept aware scenario using the UNSW dataset adopted to continual anomaly detection. The scenario includes VAE model and Replay strategy.
- Energy Dataset Example - This example showcase how to run a concept aware scenario using the Energy dataset adopted to continual anomaly detection. The scenario includes LocalOutlierFactor model and Cumulative strategy.
- Plot Heatmap Example - This example showcases how to generate a ROC-AUC heatmap for the results of a concept aware scenario
- Models examples - This directory contains multiple examples of how to use the models available in pyCLAD. Each example is a standalone script that demonstrates how to use a specific model with a specific strategy and dataset.