Back to search

Institutes outside Greece

Logic and neurosymbolic approaches for temporal phenomena recognition and multi-modal classification

Abstract

dc:description

This thesis deals with two main topics. These are, first, the use of temporal logics for temporal phenomena recognition, and second, a neurosymbolic approach for multi-modal classification. Both of these topics are motivated by real existing problems of the maritime domain. Concerning the first topic, we formally introduce a new temporal logic that was created with the aims of being very expressive without sacrificing processing efficiency and also be intuitive enough so that non-experts can use it. Our temporal logic is capable of representing instantaneous and durative temporal phenomena. We illustrate its expressive power both through examples, but also study it theoretically in comparison with other well known temporal logics. As our goal is the use of this logic for temporal phenomena recognition, we define criteria for proper stream processors, introduce our open-source implementations and evaluate them on the task of maritime activity recognition on real maritime data. Finally, we present and evaluate a methodology for extracting phenomena definitions, expressed in our logic. Concerning the second topic, that is the neurosymbolic approach for multi-modal classification, we present a methodology for learning rules, fuzzify them and integrate them into a neural network architecture. Our methodology deals with two kinds of inputs, images and related tabular data. We evaluate our approach on the task of ship type classification using real-world maritime data and vessel images. Finally, we present two maritime applications that employ the two main contributions of this thesis, i.e., our logic-based methodology for phenomena recognition, and our neurosymbolic approach for multi-modal classification. More specifically we employ them for the tasks of near-port maritime surveillance, and joint ship type classification with maritime activity recognition.

Degree

thesis:*
Grantor dc:publisher
Institutes outside Greece
Year dc:date
2024

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Pitsikalis, Emmanouil
  • Πιτσικάλης, Εμμανουήλ

Subjects

dc:subject × 22

Rights

Language dc:language
eng

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/59948

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Pitsikalis, Emmanouil; Πιτσικάλης, Εμμανουήλ. Logic and neurosymbolic approaches for temporal phenomena recognition and multi-modal classification. Institutes outside Greece, 2024. http://hdl.handle.net/10442/hedi/59948