Resilience to the Flowing Unknown: An Open Set Recognition Framework for Data Streams

Marcos Barcina-Blanco*, Jesus L. Lobo, Pablo Garcia-Bringas, Javier Del Ser

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Modern digital applications extensively integrate Artificial Intelligence models for automated decision-making. However, these AI-based systems encounter reliability and safety challenges when handling continuous data streams in dynamic scenarios. This work explores the concept of resilient AI systems that must operate in the face of unexpected events and unseen patterns. This is a common issue that regular closed-set classifiers encounter in streaming scenarios, as they are designed to compulsory classify any new observation into one of the training patterns (i.e., the so-called over-occupied space problem). In batch learning, the Open Set Recognition field addresses this issue by requiring models to maintain classification performance when processing unknown patterns. This work investigates the application of an Open Set Recognition framework that combines classification and clustering to address the over-occupied space problem in streaming scenarios. We devise a benchmark comprising different classification datasets with varying ratios of known to unknown classes, and experiments compare the performance of the proposed framework with that of individual incremental classifiers. Discussions held over the obtained results highlight situations where the framework performs best and the limitations of incremental classifiers in open-world streaming environments.

Original languageEnglish
Title of host publicationHybrid Artificial Intelligent Systems - 19th International Conference, HAIS 2024, Proceedings
EditorsHéctor Quintián, Esteban Jove Pére, José Luis Calvo Rolle, Emilio Corchado, Alicia Troncoso Lora, Francisco Martínez Álvarez, Hilde Pérez García, Francisco Javier Martínez de Pisón, Pablo García Bringas, Álvaro Herrer, Paol Fosci
PublisherSpringer Science and Business Media Deutschland GmbH
Pages139-152
Number of pages14
ISBN (Print)9783031741821
DOIs
Publication statusPublished - 2025
Event19th International Conference on Hybrid Artificial Intelligence Systems, HAIS 2024 - Salamanca, Spain
Duration: 9 Oct 202411 Oct 2024

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume14857 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference19th International Conference on Hybrid Artificial Intelligence Systems, HAIS 2024
Country/TerritorySpain
CitySalamanca
Period9/10/2411/10/24

Keywords

  • Incremental learning
  • Open set recognition
  • Open-world learning
  • Unknown classes
  •  Data streams

Fingerprint

Dive into the research topics of 'Resilience to the Flowing Unknown: An Open Set Recognition Framework for Data Streams'. Together they form a unique fingerprint.

Cite this