Μεταπτυχιακές Διατριβές

Μόνιμο URI για αυτήν τη συλλογήhttps://dspace.library.tuc.gr/handle/123456789/121

Νέα

Περιηγούμαι

Τώρα δείχνει 1 - 20 από 41

A bayesian personalized recommendation system
(Πολυτεχνείο Κρήτης, 2014) Babas Konstantinos; Μπαμπας Κωνσταντινος; Chalkiadakis Georgios; Χαλκιαδακης Γεωργιος; Lagoudakis Michael; Λαγουδακης Μιχαηλ; Deligiannakis Antonios; Δεληγιαννακης Αντωνιος
In this research, we propose a novel Bayesian approach for personalized recommendations. We succeed in providing recommendations that are entirely personalized, based on a user’s past item “consumptions”, building a representative user model which reflects agent’s corresponding beliefs. Having a set of items, our agent has to select the one which better matches her beliefs about a specific user, in order to recommend it and receive the corresponding reward. In our approach, we model both user preferences and items under recommendation as multivariate Gaussian distributions; and make use of Normal-Inverse Wishart priors to model the recommendation agent beliefs about user types. We interpret user ratings in an innovative way, using them to guide a Bayesian updating process that helps us both capture a user’s current mood, and maintain her overall user type. We produced several variants of our approach, and applied them in the movie recommendations domain, evaluating them on data from the MovieLens dataset. We developed a generic & domain independent system, able to face the scalability challenge and able to capture user preferences (long-term and short-term). Moreover, we dealt with the exploration vs exploitation dilemma in this domain, via the application of various exploration algorithms (e.g., VPI exploration). Ours is a completely personalized approach, which exploits Bayesian Reinforcement Learning in order to recommend an item or a top-N group of items, without the need of ratings prediction. We do not employ a Collaborative Filtering or Content-based or Preference Elicitation technique, but we are still able to provide successful recommendations. Furthermore, we tackle the famous “cold-start” problem via the use of Bayesian and VPI explorations. Our algorithms are shown to be competitive against a state-of-the-art method, which nevertheless requires a minimum set of ratings from various users to provide recommendations --- unlike our entirely personalized approach.
BucDoop: Bottom Up Computation of Iceberg Data Cubes with Hadoop
(Πολυτεχνείο Κρήτης, 2014) Tsakonas Konstantinos; Τσακωνας Κωνσταντινος; Deligiannakis Antonios; Δεληγιαννακης Αντωνιος; Garofalakis Minos; Γαροφαλακης Μινως; Christodoulakis Stavros; Χριστοδουλακης Σταυρος
Big Data analysis has been a key matter during the recent years for the study of various phenomena in various science contexts as well as in business intelligence. Furthermore it appears for good reason to remain in focus for the future. Online Analytical processing methods and Data Cubes need to be further studied in order to reduce time used for efficient data analysis. This study introduces BucDoop, a novel algorithm that exploits the parallelism benefits of Hadoop Map Reduce, for the efficient iceberg data cube creation in reasonable time. BucDoop includes the use of the Bottom Up Computation (BUC) idea in the context of iceberg cube data lattice traversal, managing to reduce the amount of data handled with early pruning architecture and producing the portion of the cube needed for analysis purposes (iceberg problem). Experiments conducted herein present an efficient scalability factor for the creation of the iceberg cube for very big data, by-passing the data explosion and memory constraints problem while using only commodity hardware.
Channel coding and detection for increased range bistatic scatter radio
(Technical University of Crete, 2014) Alevizos Panagiotis; Αλεβιζος Παναγιωτης; Bletsas Aggelos; Μπλετσας Αγγελος; Karystinos Georgios; Καρυστινος Γεωργιος; Deligiannakis Antonios; Δεληγιαννακης Αντωνιος
Scatter radio, i.e communication by means of reflection, has emerged as a potential key-enabling technology for ultra low-cost, large-scale, ubiquitous sensor networking. This work jointly studies noncoherent detection and channel coding for scatter radio networks, with the ultimate goal to further extend communication range, compared to prior art. Specifically, this work focuses on frequency shift keying modulation (FSK) in bistatic scatter radio architectures, where carrier emitter is dislocated from the software defined radio receiver (SDR). FSK is ideal for the power limited regime and allows for simple, frequency division multiple access (FDMA) of simultaneously operating receiver-less sensors. A novel composite hypothesis testing decoding rule is derived for noncoherent channel-encoded FSK, in bistatic scatter radio architectures. Such decoding rule is evaluated with short block length channel codes; the latter offer ultra-low encoding complexity, and thus, they are appropriate for resource-constraint scatter radio sensors. Reed-Muller and BCH codes are studied, due to their strong algebraic structure. It is shown that the proposed decoding scheme achieves high diversity order through interleaving. Extensive simulations under Rician fading scenarios include the impact of carrier frequency offset estimation errors, channel coherence time and interleaving depth. Closed-form performance analysis is also provided. Theoretical analysis for maximum likelihood coherent detection and decoding in on-off keying modulation (OOK) is also presented. Furthermore, experimental measurements are conducted outdoors, with a commodity SDR reader and custom scatter radio sensor. Sensor-to-reader ranges up to 134 meters are experimentally demonstrated with omnidirectional antennas and 13 dBm (20 milliWatt) transmission power. Coded setup offered 10 additional meters range extension compared to the state-of-the-art uncoded noncoherent detection. As a result, this thesis provided a simple solution that could further leverage the adoption of scatter radio in large-scale, ultra low-cost wireless sensor networks
Coherent detection and channel coding for backscatter sensor networks
(Πολυτεχνείο Κρήτης, 2014) Fasarakis-Chilliarnt Nikos; Φασαρακης-Χιλλιαρντ Νικος; Bletsas Aggelos; Μπλετσας Αγγελος; Karystinos Georgios; Καρυστινος Γεωργιος; Deligiannakis Antonios; Δεληγιαννακης Αντωνιος
Compact modelling of multi-gate MOSFETs for analog design
(Πολυτεχνείο Κρήτης, 2015) Gyroukis Georgios; Γυρουκης Γεωργιος; Bucher Matthias; Bucher Matthias; Balas Costas; Μπαλας Κωστας; Kalaitzakis Kostas; Καλαϊτζακης Κωστας
Τα τελευταία χρόνια, η ραγδαία ανάπτυξη της τεχνολογίας ολοκληρωμένων κυκλωμάτων, οδήγησε στην ανάγκη για μείωση του μήκους της πύλης του τρανζίστορ MOS. Η σμίκρυνση τωνδιαστάσεων των συμβατικών τρανζίστορ τεχνολογίας CMOS αναμένεται να γίνεται ολοένα και πιο δύσκολη εξαιτίας της επίπτωσης των φαινομένων μικρού καναλιού, τα οποία διαδραματίζουν όλο και σημαντικότερο ρόλο στην απόδοση της διάταξης. Τα MOSFET πολλαπλών πυλών (multi-gate MOSFET) παρουσιάζονται να είναι οι πιο ελπιδοφόρες διατάξεις που παρέχουν την δυνατότητα επέκτασης της τεχνολογίας σε διαστάσεις νάνο-κλίμακας. Αυτό οφείλεται στο γεγονός ότι λόγο του καλύτερου ηλεκτροστατικού ελέγχου του καναλιού τα φαινόμενα μικρού μήκους καναλιού καταστέλλονται. Βασικός στόχος της παρούσας μεταπτυχιακής διατριβής είναι η ανάπτυξη συμπαγών μοντέλων στα οποία θα περιέχονται εξισώσεις για το ρεύμα απαγωγού και τις διαχωρητικότητες των τρανζίστορ πολλαπλών πυλών και πιο συγκεκριμένα ενός τρανζίστορ FinFET που υπό ορισμένες συνθήκες μπορεί να αντιμετωπιστεί ως ένα τρανζίστορ διπλής πύλης (double-gate MOSFET). Τα μοντέλα αυτά θα πρέπει να μπορούν να περιγράφουν την συμπεριφορά των διατάξεων αυτών σε όλες τις περιοχές λειτουργίας. Δηλαδή από την ασθενή έως και την ισχυρή αναστροφή, κάτω και πάνω από την τάση κατωφλίου, καθώς και από γραμμική περιοχή έως την περιοχή κορεσμού. Ο τρόπο που τα περιγράφουν θα πρέπει να είναι τέτοιος ώστε να παραμένει υπολογιστικά αποδοτικό και αξιόπιστο κατά την προσομοίωση κυκλωμάτων. Επίσης να καλύπτει όλα τα φαινόμενα που προκύπτουν από την υποκλιμάκωση των δομών αυτών και τις διαφορετικές τεχνικές που χρησιμοποιούνται για την κατασκευή. Τέλος θα πρέπει να διατηρεί την απλότητα και την ακρίβεια του. Τέλος, τα μοντέλα αυτά θα πρέπει να είναι κατάλληλα δομημένα έτσι ώστε να μπορούν να εισαχθούν στα σύγχρονα περιβάλλοντα προσομοίωσης, επιτρέποντας τον σχεδιασμό CMOS κυκλωμάτων νάνο-κλίμακας. Στα συμπαγή μοντέλα, που βασίζονται στον υπολογισμό των φορτίων που εμφανίζονται στην διάταξη (charge based compact models), ρόλο κλειδί αποτελεί η σχέση που συνδέει τα φορτία που εμφανίζονται στην διάταξη με το δυναμικό που εφαρμόζεται στους ακροδέκτες αυτής. Αποτελεί λοιπόν κομβικό σημείο η όσο το δυνατόν πληρέστερη μοντελοποίηση της. Στην παρούσα διατριβή γίνεται μελέτη της συγκεκριμένης σχέσης, κάποιων απλοποιημένων μορφών αυτής και του σφάλματος που εισάγεται στο μοντέλο από τις απλουστευμένες μορφές της. Συνεχίζοντας, στην παρούσα διατριβή μελετάται η εξάρτηση της ηλεκτρικής συμπεριφοράς των νάνο-τρανζίστορ ως προς τη φυσική σχεδίαση (layout) της διάταξης. Πέρα από τις εξαρτήσεις της διάταξης από τα μεγέθη μήκος, πλάτος (και ύψος διαύλου σε FinFET τρανζίστορ), και αριθμός δακτυλίων (fingers) υπάρχει και σημαντική εξάρτηση ως προς τις αποστάσεις από μονωτικές περιοχές, την ύπαρξη dummy (ψεύδο-τρανζίστορ) κλπ. καθώς και η ανάπτυξη κατάλληλων υποκυκλωμάτων τα οποία θα αντιπροσωπεύσουν στις στατιστικές διακυμάνσεις – τόσο σε επίπεδο τεχνολογίας (process) όσο και σε επίπεδο ταιριάσματος στοιχείο-προς-στοιχείο (device-to-device matching) – συναρτήσει των γεωμετρικών δεδομένων (layout). Τέλος, περιγράφεται και η συμπεριφορά του θορύβου των τρανζίστορ, ιδίως όσων αφορά στο θερμικό θόρυβο. Τελικό αποτέλεσμα της όλης μεταπτυχιακής αυτής διατριβής είναι η παράθεση ενός Verilog-Α κώδικα για την προσομοίωση κυκλωμάτων νάνο-τρανζίστορ πολλαπλών πυλών. Οι παράμετροι του μοντέλου προσαρμόστηκαν σε δεδομένα τύπου TCAD καθώς και σε εργαστηριακές μετρήσεις.
Crowdsourcing and management of nature observational data
(Πολυτεχνείο Κρήτης, 2014) Skevakis Giannis; Σκευακης Γιαννης; Christodoulakis Stavros; Χριστοδουλακης Σταυρος; Mania Aikaterini; Μανια Αικατερινη; Deligiannakis Antonios; Δεληγιαννακης Αντωνιος
Observations of plants and animals in nature is highly valued information for the experts in the area of biodiversity. They can be used to define changes in the population of animals, plants, or track their movements throughout long periods of time. Moreover, the richer the information following the observations, the more knowledge can be extracted from them. However, the limited number of experts and the limited funding in the area, makes the observation gathering procedure almost impossible. We present the design and implementation of a framework for the management of bio- diversity observations captured by users roaming in the nature. This aims to alleviate the need for experts capturing biodiversity information, and propagates the collection of information to simple users wandering in the nature. Our framework consists of a model supporting the observations, and an infrastructure that allows the capturing, enrichment and storage of the observations using state- of-the-art technologies. Our architecture provides a scalable, highly efficient management of the collected data. The collection of the observational data is performed in real-time using mobile de- vices that most of the people have available with them, like mobile phones and tablets. Additionally, we describe the meta-model that we have defined, allowing the personalization of the metadata that follow the observations. This provides our framework with the freedom and extensibility needed so as to be implemented for various domains other than biodiversity. Finally, we describe the process of migrating the data collected by the Natural Europe project to our infrastructure.
Data integration approaches for supporting retrieval of medical information in the Web
(Πολυτεχνείο Κρήτης, 2015) Andrianakis Stamatios; Ανδριανακης Σταματιος; Petrakis Evripidis; Πετρακης Ευριπιδης; Lagoudakis Michael; Λαγουδακης Μιχαηλ; Samoladas Vasilis; Σαμολαδας Βασιλης
In recent years, the World Wide Web has become the basic source of scientific information. Especially for medical information, several Medical Information Systems have been developed in order to store and organize medical data and make it available to users through specialized search engines provided by them, or through general purpose search engines like Google. Vast amount of medical data available in the Web and the large number of Medical Information Systems make the search process of reliable medical information a time-consuming and sometimes difficult process. This work presents an integration method for search and retrieval of medical data and MIIDLE, an integration system for search and retrieval of medical information from heterogeneous sources. MIIDLE utilizes MeSH, the National Library of Medicine's controlled vocabulary thesaurus as a common vocabulary for the integration process. Using AMTEx method, it extracts medical terms from the retrieved data, and it expands the query used for the retrieval with MeSH terms. Combining the extracted terms with the expanded query it ranks the results with respect to their relevance using Vector Space Model. MIIDLE results are evaluated by users and its performance is compared with the performance of the sources that it accesses.
Depth perception from a single camera and multiple light sources
(Technical University of Crete, 2015) Rematska Georgia; Ρεματσκα Γεωργια; Dollas Apostolos; Δολλας Αποστολος; Papaefstathiou Ioannis; Παπαευσταθιου Ιωαννης; Zervakis Michalis; Ζερβακης Μιχαλης
3D Vision has always been a subject of research for many decades. The use of images results in the loss of the 3rd dimension. Many techniques have been developed over the years that aim in acquiring depth from images. Among the most widespread are stereo vision, fringe projection, laser scanning or combinations of them, all of which have shown high quality results but unsuitable for some classes of applications. For example, laser scanning rangefinders are not low-cost system. Hence, there is a need to develop a low cost, high speed, high resolution system for 3D vision. In this thesis a novel approach is presented to estimate depth using a single camera combined with two spectrally distinct light sources for road surface measurements. The light sources consist of two sets of LED arrays. Depth information can be extracted, by processing the different reflections in the image from the two light sources. The basic information of depth estimation lies in their blue to red ratio, which with appropriate calibration is correlated to depth. The methodology was first developed in MATLAB and then was implemented on FPGA. The system is real time and fully optimized to have the minimum hardware resources and high frequency of operation for high resolution images.
Design and layout techniques in analog/RF integrated circuits
(Technical University of Crete, 2014) Papathanasiou Konstantinos; Παπαθανασιου Κωνσταντινος; Bucher Matthias; Bucher Matthias; Kalaitzakis Kostas; Καλαϊτζακης Κωστας; Balas Costas; Μπαλας Κωστας
Τα τελευταία χρόνια, η ραγδαία ανάπτυξη στα αναλογικά ολοκληρωμένα κυκλώματα πολύ υψηλών συχνοτήτων (RFIC) οφείλεται κυρίως στην εξέλιξη της τεχνολογίας CMOS. Οι λόγοι που οδήγησαν σε αυτήν την ανάπτυξη συνοψίζονται κυρίως στους εξής: την υψηλή απόδοση, που απορρέει από τη σμίκρυνση του μήκους καναλιού των CMOS τεχνολογιών, το χαμηλότερο κόστος τους, συγκριτικά με άλλες τεχνολογίες, καθώς και την γρήγορη ενσωμάτωση τους στον τομέα της μικροηλεκτρονικής. Αποτέλεσμα των παραπάνω, είναι τεχνολογίες CMOS με μήκος καναλιού κάτω από τα 45nm να θεωρούνται, πλέον, αιχμή της τεχνολογίας. Λόγω του υψηλού λόγου απόδοσης προς κόστος και της αξιοπιστίας (reliability) τους, τεχνολογίες με μεγαλύτερα μήκη καναλιού (90nm ή και 180nm) χρησιμοποιούνται ακόμα και στις μέρες μας κατά κόρον στην σχεδίαση και κατασκευή RFICs. Σκοπός της παρούσας μεταπτυχιακής εργασίας είναι η μελέτη και αξιοποίηση τεχνικών σχεδίασης για RFICs υψηλής απόδοσης, καλύπτοντας ένα μεγάλο εύρος CMOS τεχνολογιών, από τα 90nm ως τα 30nm. Η παρούσα έρευνα εκτείνεται από την ορθή υλοποίηση του φυσικού σχεδίου (layout) RF διατάξεων και κυκλωμάτων έως τον καθορισμό του βέλτιστου σημείου λειτουργίας τους και βρίσκει εφαρμογή σε κυκλώματα ενισχυτών χαμηλού θορύβου (LNA). Για τον σκοπό αυτό σχεδιάστηκε, υλοποιήθηκε και κατασκευάστηκε ένα RF τεστ τσιπ σε τεχνολογία 90nm της TSMC. Οι δομές που υλοποιήθηκαν, στη συνέχεια μετρήθηκαν on wafer,, χαρακτηρίσθηκαν και μοντελοποιήθηκαν με το EKV3 μοντέλο έως τα 26.5GHz. Εν συνεχεία, μελετήθηκε και παρουσιάζεται η επίδραση βασικών παραμέτρων σχεδίασης, όπως μήκος (L), πλάτος (W) καναλιού, αριθμός δακτύλων (NF) στην απόδοση των MOSFET δομών και των ενισχυτών χαμηλού θορύβου, μέσω αντιπροσωπευτικών δεικτών απόδοσης (figures of merit), σε τεχνολογία με μήκος καναλιού 90nm. Οι συγκεκριμένοι δείκτες απόδοσης μελετήθηκαν επίσης σε προηγμένη CMOS τεχνολογία με μήκος καναλιού 30nm. Tα αποτελέσματα, τα οποία επικυρώθηκαν με το EKV3 μοντέλο, ανέδειξαν την μετατόπιση του βέλτιστου σημείου λειτουργίας RF κυκλωμάτων προς το μέσο της περιοχής μέτριας αναστροφής, με την μείωση του μήκους καναλιού. Το γεγονός αυτό είναι πολύ σημαντικό, καθώς καθιστά εφικτή την μείωση της κατανάλωσης ισχύος, με ταυτόχρονη αύξηση της συνολικής απόδοσης των εν λόγω κυκλωμάτων.
Design and performance evaluation of sensing algorithms and cooperative relay selection protocols for multichannel cognitive radio networks
(Technical University of Crete, 2015) Theodorou Maria; Θεοδωρου Μαρια; Paterakis Michalis; Πατερακης Μιχαλης; Koutsakis Polychronis; Κουτσακης Πολυχρονης; Liavas Athanasios; Λιαβας Αθανασιος
Recent years have witnessed a dramatic increase in the demand for radio spectrum. This is partly due to the increasing interest of consumers in wireless services, which in turn is driving the evolution of wireless networks toward high-speed data networks. Cognitive radio has been proposed as a promising technology to improve the spectral efficiency of radio spectrum, and is achieved by allowing unlicensed secondary users (SUs) to coexist with licensed primary users (PUs) in the same spectrum. The primary network owns the spectrum, and has performance guarantees. The secondary network(s) can access the spectrum if no significant degradation on the primary communication is caused. In this Thesis we start in Chapter 2 by proposing four new transmission algorithms for multichannel homogeneous cognitive radio networks (CRNs). We examine two cases: (i) the case where the network’s channels are not assigned to the SUs by a centralized entity and (ii) the case where a centralized entity exists and assigns the network’s channels to the SUs. Our event-driven simulations results demonstrate that the new transmission algorithms we have introduced improve (i) the normalized average throughput of SUs, (ii) reduce the dropping probability and (iii) increase the number of successful transmissions occurring during the system operation, when compared with a popular algorithm proposed in recent work in this area. Chapter 3 of the Thesis studies new transmission algorithms for multichannel heterogeneous CRNs. As in the first part we examine two cases: (i) a distributed CRN and (ii) a centralized CRN. For each case and for the same network topology, as in the first part of the work, we propose a new algorithm. Our event-driven simulations results demonstrate that the new transmission algorithms we have introduced considerably improve the average number of Mbits of secondary user traffic transmitted in each time slot, when compared with the corresponding results of the “γ-persistent strategy” recently introduced in the literature. In Chapter 4 of the Thesis the “Distributed algorithm” proposed in Chapter 3, in which the SUs select their network’s channels in a distributed way without coordination by a centralized entity, is used and evaluated in the case of homogeneous multichannel CRNs. Our simulation results demonstrate that the “Distributed algorithm” achieves results close to those achieved by the algorithms proposed in Chapter 2 of the Thesis in which it has been assumed that a centralized entity exists and assigns the network’s channels to the SUs. Finally, in Chapter 5 of the Thesis new cooperative communication protocols are proposed for cognitive radio networks, in which one primary user and multiple SUs cooperate for mutual benefit. We proposed and evaluate two new protocols, the Best Relay Selection Protocol (BRSP) and the Stopping Criterion Protocol (SCP) which allow cooperation between the PU and the SUs. Our simulation results demonstrate that the proposed protocols decrease the total primary packet transmission time, compared with the time required for direct packet transmission by the PU.
Design optimization of an electric energy production system for power supplying the nodes of wireless sensor networks
(Technical University of Crete, 2015) Mandourarakis Ioannis; Μανδουραρακης Ιωαννης; Koutroulis Eftychios; Κουτρουλης Ευτυχιος; Kalaitzakis Kostas; Καλαϊτζακης Κωστας; Stavrakakis Georgios; Σταυρακακης Γεωργιος
The widespread utilization of the Wireless Sensor Networks (WSNs) that are power supplied by Renewable Energy Sources (RES) is now leading scientific research towards the development of innovative sizing optimization techniques and configurations that aim in the service of multi-criteria objectives of economical and/or technical nature. The most frequently used ones have to do with the minimization of the overall cost and the maximization of the overall system efficiency, in terms of energy redundancy and operational reliability. Sizing optimization techniques are being introduced, where there is a tradeoff amongst similar to the aforementioned criteria that often contradict to each other (e.g. reliability is improved when economical cost or energy redundancy are increased). In this thesis, two complementary design optimization methods (a circuit level and a system level study) are presented for deriving the optimal configuration of the RES based energy production system of a WSN node, such that its total lifetime cost is minimized, while simultaneously guaranteeing that the data acquisition equipment is uninterruptedly power supplied during the entire year. The experimental results verify that, by applying the design variables as they were derived by the proposed optimization techniques at both the circuit and the system level, RES-based power-supply structures with a lower lifetime cost and higher power-processing efficiency are derived, compared to the non-optimally designed configurations. The design optimization and experimental results indicate that by using the proposed techniques, the total cost of the RES-based power supply system is reduced by 15.7 % and the DC DC converter efficiency is increased by 5.5 % compared to the corresponding results obtained by non-optimized power-supply structures.
Full-rate fifferential M-PSK Alamouti modulation with polunomial-complexity maximum-likelihood noncoherent detection
(Technical University of Crete, 2012) Markopoulos Panagiotis; Μαρκοπουλος Παναγιωτης; Karystinos Georgios; Καρυστινος Γεωργιος; Liavas Athanasios; Λιαβας Αθανασιος; Bletsas Aggelos; Μπλετσας Αγγελος
Grammatical inference for event recognition
(Πολυτεχνείο Κρήτης, 2014) Kofinas Nikolaos; Κοφινας Νικολαος; Lagoudakis Michael; Λαγουδακης Μιχαηλ; Bletsas Aggelos; Μπλετσας Αγγελος; Garofalakis Minos; Γαροφαλακης Μινως
As robot technology finds applications in the real world (search and rescue, daily household tasks, etc.), huge amounts of data are generated during autonomous robot missions. In such applications, it is often desirable to recognize high-level events that may have occurred during a mission either online or offline. Event Recognition in robot missions currently relies on human expertise and time-consuming data annotation. A modern method to recognize events is to employ Probabilistic Context-Free Grammars (PCFGs), which are formal models that can capture complex patterns in discrete sequences and can be used to parse incoming sensor data streams in order to detect patterns that may signal the occurrence of some event of interest. Recent experimentation with such methods on data from Autonomous Underwater Vehicle (AUV) missions indicated that interesting events can be recognized by parsing sequences of sensor data using an intuitive hand-written PCFG. This thesis introduces a generic procedure which can be used to automatically construct PCFGs which encode sensor data sequences that typically appear during normal robot operation using recorded logs from past missions. The resulting PCFGs can be used to recognize abnormal events in new missions evidenced by sensor data sequences which cannot be interpreted as normal. The proposed procedure consists of two parts: (a) the transformation of sensor streams into discrete sequences either to form a training corpus offline or to generate input for online parsing and (b) a Grammatical Inference algorithm in order to learn a compact PCFG consistent with a given training corpus. The learning part relies on a local search method over the space of possible grammars using chunk and merge operations. The search method aims to find a compact grammar that also maximizes its posterior probability, in a Bayesian sense, with respect to a given training corpus. The proposed procedure is evaluated on a variety of domains ranging from data-sets generated by typical context-free grammars to data-sets generated from real robot missions (NAO robot walk and AUV navigation). The results indicate that our approach is capable of producing reliable PCFG-based event recognizers, which may yield some false positive signals, but in general succeed in capturing abnormalities.
Hardware accelerated basic blocks for power-aware intercommunication in HPC and embedded systems
(Πολυτεχνείο Κρήτης, 2014) Tampouratzis Nikolaos; Ταμπουρατζης Νικολαος; Papaefstathiou Ioannis; Παπαευσταθιου Ιωαννης; Dollas Apostolos; Δολλας Αποστολος; Pnevmatikatos Dionysios; Πνευματικατος Διονυσιος
In the past, a transition to the next fabrication process typically translated to more transistors and frequency and less power. The higher frequencies paired with innovations in computer architecture defined the semiconductor industry and research until the mid-90s. At that point architecture research saturated and industry resided to the technology scaling for performance gains. During the mid-00s frequency scaling saturated as well. Transistor count, the only resource which reliably kept scaling, along with intra-chip parallelism, which could leverage and extend the existing knowledge of old-days supercomputers, emerged as the only solution to keep Moore’s law live. In parallel systems, computing nodes cooperate to solve processing intensive problems. The communication between nodes is achieved through a variety of protocols. Traditionally, research has focused on optimizing these protocols and identifying the most suitable ones per system and application. Recently, an attempt to unify the primitive operations of the proposed intercommunication protocols has been realized through the Portals system. Portals offer a set of low level communication routines which can be composed to model complex protocols. However, Portals modularity comes at a performance cost, as communication protocols have been tuned and many of their timing critical parts have been decoupled from the main execution thread and in many cases accelerated as dedicated hardware. This work targets to close the performance gap between a generic and reusable intercommunication layer, Portals, and the several monolithic but highly tuned protocols. A software driven hardware accelerated system is suggested which resides on execution of actual software to highlight the critical parts of the communication routines. Accelerating the bottlenecks starts by modeling the hardware in untimed virtual prototypes and the software in a range of candidate embedded processors. A novel path from hardware prototypes to actual silicon allows rapid characterization of the accelerator in terms of power, performance and area. The suggested approach triggers a speedup from one order of magnitude in bottleneck components of Portals, while it is up to two orders of magnitude faster in both MPI and GA baseline implementations in a recent embedded processor.
Hyper spectral data estimation from power dimensionality experimental imaging
(Technical University of Crete, 2014) Iliou Dimitrios; Ηλιου Δημητριος; Balas Costas; Μπαλας Κωστας; Garofalakis Minos; Γαροφαλακης Μινως; Digalakis Vasilis; Διγαλακης Βασιλης
We report the first real time modular spectral and color imaging system based on the combination of snapshot spectral imaging, spectral estimation and color reproduction algorithms. A limited number of spectral bands are captured simultaneously, with the aid of specially designed camera, which are subsequently processed with spectral estimation algorithms to obtain a full spectrum per image pixel. We have succeeded to demonstrate complete spectral cube calculation and display of millions of spectra in real-time and to remove trade-off between spectral and spatial resolution. Besides accurate spectral mapping, our approach enables also reliable and device-independent color reproduction based on complete, per-pixel spectra. These achievements hold the promise to provide an indispensable tool in nondestructive analysis and in noninvasive diagnosis.
Machine learning methods for genomic signature extraction
(Technical University of Crete, 2015) Chlis Nikolaos-Kosmas; Χλης Νικολαος-Κοσμας; Zervakis Michalis; Ζερβακης Μιχαλης; Balas Costas; Μπαλας Κωστας; Mania Aikaterini; Μανια Αικατερινη
The application of machine learning methodologies for the analysis of DNA microarray data has become a common practice in the field of bioinformatics. DNA microarrays can be used in order to simultaneously measure the expression value of thousands of genes. Given the measurements of gene expression, machine learning methods can be employed in order to identify candidate genes that are related to a biological state or phenotype of interest, such as cancer. These lists of candidate genes are often called “genomic signatures” in literature. The application of machine learning methods for the extraction of genomic signatures is a necessity, since it is practically impossible for field experts to assess the importance of each gene individually by manual inspection due to the large size of the genome, which consists of approximately 25,000 genes. Machine learning methods such as feature subset selection and classification algorithms are popular choices for the extraction of genomic signatures. Univariate feature selection methods filter genes according to difference in their gene expression profiles among samples belonging to different classes of interest, such as control and disease. Since they test each gene individually, univariate methods are computationally efficient and they select genes with high discrimination ability. However, they ignore associations among genes. On the other hand, multivariate methods simultaneously assess groups of genes and select candidate genes based on their predictive performance when used in conjunction with a classifier. As such, they are more efficient at capturing the latent associations among genes and select genes with high predictive capability, at the cost of being computationally expensive. While the applied feature selection and classification methodologies have matured and several state of the art algorithms have been established, the stability of the extracted genomic signatures is often overlooked. As a result, the genomic signatures extracted by many methodologies are unstable under sample variations. That is, the extracted signatures differ significantly under variations of the training data. Since result stability is related to generalization, this instability raises skepticism in the expert community and hinders the validity and clinical application of research findings extracted from such gene expression studies. This thesis deals with the following three aspects of the selection and evaluation of gene signatures, namely stability, predictive capability and statistical significance. First, a framework for the extraction of stable genomic signatures, called Stable Bootstrap Validation (SBV) is introduced. The proposed methodology enforces stability at the validation step. As a result, it can be combined with any classification method, as long as it supports feature selection. Three publicly available gene expression datasets are used in order to test the proposed methodology. First the dimensionality of the datasets is reduced using a filtering method. Then, bootstrap resampling is utilized in order to generate a list of candidate signatures according to the selection frequency of genes across all bootstrap datasets. Then, a stable signature which has maximal predictive performance in terms of accuracy, sensitivity and specificity is extracted and the predictive performance of all candidate signatures is plotted in an elaborate manner for further inspection. Additionally, the application of random sampling methods for countering the negative effects of imbalanced datasets in classification was investigated, since imbalanced datasets are frequently found in DNA microarray studies where control samples are usually scarce. Moreover, a proper statistical framework was implemented that includes two separate statistical tests, in order to assess the statistical significance of the extracted signature in terms of classification accuracy as well as association to the response variable (phenotype/biological state). Finally, the robustness of the methodology is assessed by testing the degree of “agreement” among signatures extracted from independent executions of the methodology.
Non-coherent receivers for zero-feedback distributed beamforming in connectivity-constrained wireless sensor networks (WSNs)
(Πολυτεχνείο Κρήτης, 2014) Alexandris Konstantinos; Αλεξανδρης Κωνσταντινος; Bletsas Aggelos; Μπλετσας Αγγελος; Karystinos Georgios; Καρυστινος Γεωργιος; Liavas Athanasios; Λιαβας Αθανασιος
Power-constrained wireless sensor networks (WSNs) suffer from network partitioning problems. In many cases, each node among a network subset, cannot reliably communicate with a distant receiver even when transmitting at maximum power. Thus, a collaborative beamforming scheme among the distributed adjacent terminals is needed in terms of power addition. Prior art on distributed beamforming has mainly focused on feedback messages for channel estimation (CSI) or physical layer carrier phase adjustments. In sharp constrast, this thesis assumes commodity radios and studies the low signal to-noise-ratio (SNR) regime, where accurate channel estimation is not feasible and no reliable feedback exists. The main idea is to exploit recently proposed zero-feedback distributed beamforming and design specific non-coherent receivers. Towards that goal, three concrete non-coherent receivers are presented for zero-feedback distributed beamforming (ZF-DBF); one based on energy detection, one based on maximum-likelihood for a specific condition (i.e., full correlation among the received samples), and finally, one non-coherent receiver for all other cases. A non-coherent receiver for energy harvesting through time division multiple access (TDMA) is also provided for comparison purposes. Analytical and numerical bit-error-rate results are presented. It is shown that the ZF-DBF receiver outperforms the energy harvesting one at the low-SNR regime and overcomes connectivity adversities by exploiting signal alignment from the distributed terminals, at the expense of total (network) power transmission.
Non-linear synchronization methods on magnetoencephalographic (MEG) recordings
(Technical University of Crete, 2015) Antonakakis Marios; Αντωνακακης Μαριος; Zervakis Michalis; Ζερβακης Μιχαλης; Lagoudakis Michael; Λαγουδακης Μιχαηλ; Mania Aikaterini; Μανια Αικατερινη
Cross-frequency coupling (CFC) is thought to represent a basic mechanism of functional integration of neural networks across distant brain regions. Furthermore, several neuroimaging studies have suggested that functional brain connectivity networks exhibit “small-world” characteristics, whereas recent studies based on structural data have proposed a “rich-club” organization of brain networks, whereby nodes of high connection density tend to connect among themselves compared to nodes of lower density. In this study, CFC profiles are analyzed from resting state Magnetoencephalographic (MEG) recordings obtained from 30 mild traumatic brain injury (mTBI) patients and 50 controls. The non-linear synchronization metric, mutual information (MI) is used to quantify the phase-to-amplitude coupling (PAC) of activity among the recording sensors in six nonoverlapping frequency bands. After forming the CFC-based functional connectivity graphs (FCGs), a tensor representation and tensor subspace analysis is employed to identify an set of features with low dimensions for subject classification as mTBI or control. Keeping FCGs from the optimal set of features, an “attack strategy” to is developed to compare the rich-club and small-world organizations and identify the model that describes best the topology of brain connectivity. Results show that the controls form a dense network of stronger local and global connections, indicating higher functional integration compared to mTBI patients. Furthermore, mTBI patients could be separated from controls with more than 90% classification accuracy. Finally, the results suggest that resting state MEG connectivity networks follow a rich-club organization. These findings indicate that the analysis of brain networks computed from resting-state MEG with PAC and tensorial representation of connectivity profiles may provide a valuable biomarker for the diagnosis of mTBI.
Optimal blind detection of APSK in polynomial time
(Technical University of Crete, 2014) Fountzoulas Ioannis; Φουντζουλας Ιωαννης; Karystinos Georgios; Καρυστινος Γεωργιος; Bletsas Aggelos; Μπλετσας Αγγελος
Oλοκληρωμένο σύστημα μέτρησης φθοράς κοπτικών εργαλείων με τη βοήθεια ψηφιακής επεξεργασίας εικόνας
(Technical University of Crete, 2014) Lyronis Antonios; Λυρωνης Αντωνιος; Zervakis Michalis; Ζερβακης Μιχαλης; Antoniadis Aristomenis; Αντωνιαδης Αριστομενης; Kalaitzakis Kostas; Καλαϊτζακης Κωστας

Περιηγούμαι

Πλοήγηση Μεταπτυχιακές Διατριβές ανά Τίτλο