Lectures / Seminars

Pattern mining in multi-relational data

"Pattern mining research has so far mostly focused on mining a single relation. The most representative example of such methods is frequent itemset mining which aims at discovering frequently co-occurring items from a relation of items and market basket transactions. However, most real world datasets are usually richer, containing more than one relation and more than two entity types (ex. characteristics of items).Therefore pattern mining methods that target multi-relational data are equally important.

N-gram graphs and proximity graphs: bringing summarization, machine learning and bioinformatics to the same neighbourhood



N-gram graphs offer a generic representation and a related framework for the representation and handling of symbol sequences. They encode neighbourhood information between items expressing both local and global characteristics of the sequences in a manner that supports uncertainty. They have been applied successfully to several tasks, such as news summarization (NewSum method) and news summarization evaluation (AutoSummENG, MeMoG, NPowER methods). N-gram graphs can also co-exist with and empower existing machine learning algorithms in the vector space.

Event Recognition for Unobtrusive Assisted Living

Developing intelligent systems towards automated clinical monitoring and assis-tance for the elderly is attracting significant attention. Age-related demographic trends in most western countries and increasing health-care costs, indicate a need for robust telehealth solutions which shall prolong seniors’ independent living. A key requirement in most ambient intelligence and assisted living applications is unobtrusiveness: Monitoring should not intervene with daily activities, so that the user feels comfortable and sensor data are collected naturally and in an unbiased fashion.

MBrace: large scale distributed computation

Με την έλευση των cloud platforms και των private data centers οι προγραμματιστές αντιμετωπίζουν για μια ακόμη φορά τις προκλήσεις του distributed computing στην προσπάθειά τους να αξιοποιήσουν τη διαθέσιμη υπολογιστική ισχύ. Concurrency, message passing, elasticity & machine failure είναι πλέον κοινές καταστάσεις που η ομάδα εργασίας οφείλει να λάβει υπόψη κατά την ανάπτυξη κώδικα που απαιτείται για την εκτέλεση των big computations ή την διαχείριση big data ή και τα δύο.

Incremental learning of event definitions with Inductive Logic Programming

Εvent Recognition systems rely on properly engineered knowledge bases of event definitions to infer occurrences of events in time. The manual development of such knowledge is a tedious and error-prone task, thus event-based applications may benefit from automated knowledge construction techniques, such as Inductive Logic Programming (ILP), which combines machine learning with the declarative and formal semantics of First-Order Logic.

Building Bilingual Multiword Lexicons from Parallel Text

Multiword expressions resources are important for both rule-based and statistical machine translation. We present a method to construct bilingual multiword lexicons from SMT phrase tables. The lexicons developed in the grammar formalism GF, which ensures syntactical correctness and generates all the bending forms of the entries. The resources created in this manner can be used to enrich either GF grammars or SMT phrase tables.

Εκτατικός Λογικός Προγραμματισμός Υψηλής Τάξης

Τα δύο δημοφιλή παραδείγματα δηλωτικού προγραμματισμού, δηλαδή ο λογικός και ο συναρτησιακός προγραμματισμός διαφέρουν σημαντικά: ο λογικός προγραμματισμός είναι παραδοσιακά πρώτης τάξης ενώ ο συναρτησιακός προγραμματισμός προωθεί την χρήση συναρτήσεων υψηλής τάξης.

Αλγόριθμοι εμπιστοσύνης στο σύστημα κοινωνικής δικτύωσης QSN

Το QSN αποτελεί ένα σύστημα κοινωνικής δικτύωσης που δίνει την δυνατότητα στους χρήστες του να εκφράζουν με δομημένο τρόπο τις απόψεις τους για διαδυκτακές πηγές. Το σύστημα αναλαμβάνει την συνάθροιση των διαφορετικών απόψεων που εκφράζονται, παράγοντας τελικώς ετικέτες χαρακτηρισμού περιεχομένου για κάθε χαρακτηρισμένη διαδικτυακή πηγή. Σκοπός της παρούσας εργασίας ήταν η διερεύνηση των αλγόριθμων εμπιστοσύνης, και η αξιοποίησή τους στο σύστημα QSN με στόχο την βελτίωση της ποιότητας των παραγόμενων ετικετών.

Graph-theoretic conflict resolution with applications in wireless networks and protein docking

We consider graphs with nodes representing decisions and edges representing conflicts among two decisions. Nodes are further assigned weights indicating the reward of the corresponding decision. In such a graph, the Maximum Weighted Independent Set (MWIS) problem is to select a set of nodes, no two of which are adjacent, with the largest possible total weight. This is equivalent to selecting a "conflict-free" set of decisions with maximal reward. MWIS is NP-hard.

Semi-automatic road extraction combining Particle Filtering and Geographic Information Systems

The updating of road network databases is crucial to many Geographic Information System (GIS) applications like navigation, urban planning, as well as emergency and disaster management. The development of a robust methodology for automatic or semi-automatic road extraction and change detection as well as “discovery of paths” (in military theater, desert areas etc.) is essential. Such a methodology has to provide accurate and up-to-date results albeit using noisy and infrequent sensor data.

Syndicate content

© 2019 - Institute of Informatics and Telecommunications | National Centre for Scientific Research "Demokritos"

Terms of Service and Privacy Policy