The proposed (K)-SIF method significantly enhances anomaly detection in complex data by leveraging the signature method, demonstrating robustness against noise and effective parameter sensitivity.
Numerical experiments revealed that adjusting the depth parameter is crucial for the signature isolation forest algorithms, directly impacting their performance and resilience in varying data conditions.
Comparative assessments showed that (K)-SIF outperforms traditional functional isolation forest methods (FIF) in terms of computational efficiency, which is particularly beneficial for large datasets.
In benchmarking against real-world anomaly detection tasks, (K)-SIF consistently exhibited superior discrimination power, reinforcing its viability for practical applications in singular and diverse datasets.
Collection
[
|
...
]