Der 11. Bernoulli-ims World Congress in Probability and Statistics ist die wichtigste internationale Konferenz auf dem Gebiet der Wahrscheinlichkeitstheorie und der mathematischen Statistik, auf der Wissenschaftler:innen aus der ganzen Welt Ideen austauschen und ihre neuesten Forschungsergebnisse pr\u00e4sentieren.<\/p>\n
Unsere Kollegin Dr. Marisa Mohr wird am 16. August gemeinsam mit Dr. Nils Finke (Oldendorff Carrier GmbH & Co. KG) mit dem Talk \u201eMultivariate Ordinal Pattern Representations in Real-World Applications\u201c<\/em><\/strong> vertreten sein.<\/p>\n Abstract:<\/em><\/strong> Unsere Kollegin Dr. Marisa Mohr wird am 16. August gemeinsam mit Dr. Nils Finke (Oldendorff Carrier GmbH & Co. KG) mit dem Talk \u201eMultivariate Ordinal Pattern Representations in Real-World Applications\u201c vertreten sein.<\/p>\n","protected":false},"featured_media":56673,"parent":0,"template":"","events_types":[796,795],"service":[76],"class_list":["post-56666","events","type-events","status-publish","has-post-thumbnail","hentry","events_types-kongress","events_types-vortrag","service-artificial-intelligence"],"acf":{"start_date":"20240812","finish_date":"20240816","time":null,"finish_time":null,"location":"Bochum","organizer":"","file":null,"hubspot_signups":false,"hubspot_signups_text":"","add_speakers":false,"speakers_headline":"Speaker:innen","speakers":null,"contact":null,"contact_heading":"","contact_text":""},"yoast_head":"\n
\nThe predictive capacity of machine learning models depends critically on the quality of training data. Here, feature extraction or feature learning is concerned with finding efficient mappings from raw data to (low-dimensional) representations that capture information appropriate for the learning task. Information-theoretic entropies, especially permutation entropy, hold promise for encoding intrinsic time-series patterns and facilitating efficient mappings for predictive modeling. Time series representations based on information theoretic entropies are a proven and well-established approach. Since this approach assumes a total ordering it is only directly applicable to univariate time series and thus rendering it difficult for many real-world applications dealing with multiple measurements at the same time, i.e., multivariate time series. In this session, we present different variants of multivariate ordinal pattern permutation entropy as a new approach that takes into account the correlation between multiple variables. This approach promises several advantages, which we will show in two different application fields. We compare its effectiveness with other advanced features, e.g. other statistical measures or internal representations computed by deep neural networks. In addition, the approach is suitable to speed up inference in lifted inference in constructing temporal symmetries on large dynamic probabilistic graphical models for a dry bulk shipping company.<\/em><\/p>\n","protected":false},"excerpt":{"rendered":"