Recent advances in hardware technology have facilitated new ways of collecting data continuously, giving rise to data streams. In many applications, the volume of such data is so large that it may be impossible to store the data on disk. Furthermore, even when the data can be stored, the volume of incoming data may be so large as to be impossible to process. Internet traffic, financial transactions related to securities and options, and sensor-based data collection related to continuous physical observations such as temperature, pressure, human EEG signals or vibrations in manufacturing industries, are typical data streaming scenarios. Thus, the development of machine learning algorithms becomes significantly challenging in this context.
In our view, organizing this workshop is timely because as artificial intelligence-based decisions become more ubiquitous to individuals and society, modeling data streams is going to play a major role on the development of human-centered algorithms and machine learning models that account for the feedback loop between machine and human decisions, which are inherently asynchronous events.
Supervised classification models for data streams
Clustering methods for data streams
Discovering association methodology for data streams
Real and virtual concept drift detection methods
Evaluation measures in machine learning for data streams
Anomaly and novelty detection for data streams
Duration and format: Half-day in the morning
Boaz Lerner University of the Negev Israel Title of talk: Detection of Concept Drift, Novelty, and Latent Mechanisms in Data Streams
The accepted papers will be published as an edited book by Now Publishers, indexed in leading databases such as the Web of Science Book Citation Index and SCOPUS. There will be the hardcover (full color) and the eBook versions. The paper with the best reviews will be invited to be extended and submitted to the JCR "International Journal of Interactive Multimedia and Artificial Intelligence" journal, in a special issue that will be published by 2021. There will be a special issue in "Progress in Artificial Intelligence" for an extended version of the second-best paper.
Papers must be formatted according to the ECAI2020 Formatting Instructions and up to 7 pages in length + 1 page references in PDF format (full papers)