Περιγραφές θέσεων για εννέα (9) υποτροφίες για την εκπόνηση διδακτορικής διατριβής του ΙΠ&Τ σε συνεργασία με Πανεπιστήμια του εξωτερικού

Printer-friendly versionSend by email

2 θέσεις για το Rice University

Τίτλος 1ης θέσης: Robot motion planning

Περιγραφή 1ης θέσης: In human-human collaboration or manual work, people are always able to use cues and behaviours to converge to an implicitly mutually agreed model for collaboration without any need to explicitly discuss and carefully choreograph their work in advance. Physical human-robot interaction on the other hand, is often formulated as a static, pre-configured model for allocating sub-tasks towards a common goal to the robot and to the human and not as a genuine and active collaboration. However, different human collaborators might be more skilled in or more keen to assume different subtasks. The human collaborators might also experiment with different task allocations and eventually converge to an allocation that is efficient and productive. Even more dynamically, fatigue, avoidance of repetition, or any number of external factors might prompt a human collaborator to change their preference on what subtasks to assume and what subtasks to leave to the robot on a daily basis.
The aim of this PhD project is to develop motion planning methods that solve constraints related to both environment states and human behaviour, integrating adaptation to and learning from each other. Some of the challenges will be to define a suitable representation of the human collaborator and to develop a new motion planning paradigm for this more complex representation that is fast, reactive, and responsive to the requirements that arise from human-robot interaction.

Τίτλος 2ης θέσης:Biological interaction modelling and analysis

Περιγραφή 2ης θέσης: Biological systems and processes are based on various types of interaction among biological entities. At the lowest level, one finds interactions among individual molecules or complexes thereof, such as protein-ligand interactions. Such elementary interactions become the building blocks of more complex pathways, such as metabolic ones, and complex biological networks associated with particular cell functions, e.g. interactions among proteins in muscle cells. The computational modelling of biological interactions has become essential in understanding biological functions at various levels of abstraction. For instance, such modelling helps predicting the binding potential among proteins and ligands, based on structural, thermodynamic or chemical properties of the entities. At a higher level of abstraction, it also helps identifying functional associations among proteins or protein complexes, based on their interactions in a corresponding network.
This PhD thesis aims to introduce novel techniques and algorithms for such computational modelling and analysis, focusing on the interactions of various molecules (e.g. RNA or proteins) with proteins. Research in this context may involve a number of cutting-edge approaches, ranging from structure-based binding prediction to hypergraph analysis methods for complex biological networks.


1 θέση για το University of Texas at Arlington

Τίτλος: Robot motion signalling in joint action

Περιγραφή: The ability of both humans and robots to understand each other and collaborate effectively and safely is essential for achieving intuitive and seamless human-robot collaboration, in applications such as intelligent manufacturing or industry 4.0: how humans can collaborate with robots in the factory of the future.
In order to achieve that the human collaborator is able to correctly interpret the robot's motion and for the robot to understand and learn from the human’s actions, we need advanced machine learning algorithms. In addition, robot motion planning is needed that can reproduce cues and signals used in human-human collaboration. Naturally, making robot motion legible and predictable by humans might run counter to other qualities, such as the efficiency and speed of the motion.
The aim of this PhD project is to develop motion planning methods that embed such cues in the motion plan, to evaluate their eligibility and predictability in user studies, and to compare their efficiency and speed against plans that only take the latter into consideration. One example use case can be providing appropriate cues in a handover motion that the robot is about to release the object, even if this reduces the object-handing throughput. Another example use case can be that a robot that needs to reposition a number of objects does this in an order that is sub-optimal, but guarantees that it is at all times obvious which object it is going for next. This PhD project can involve socially assistive robots, arm robots and different multisensing devices to interact, including virtual and augmented reality, to enable the robot to learn.


5 θέσεις για το University of Houston

Τίτλος 1ης θέσης: Identification of argument elements and argument relations in natural laguage texts

Περιγραφή 1ης θέσης: This PhD thesis aims to tackle a relatively new and exciting challenge in corpus-based discourse analysis, argument mining, which tries to automatically identify human reasoning in unconstrained natural language texts. The PhD student will be able to explore theories developed in the humanities for describing argument and rhetoric and approaches based on machine learning and statistical techniques from computational linguistics, in order to address challenges similar to the following ones:
- Define algorithms for identifying argument elements in natural language texts.
- Propose algorithms for identifying intra-argument relations, organizing the identified elements that belong to a single argument in order to mine the internal structure of an argument.
- Propose algorithms for identifying inter-argument relations, that determine the role of each argument in the narrative of a text passage.

Τίτλος 2ης θέσης: Content-based analysis of multimedia

Περιγραφή 2ης θέσης: Modern multimedia databases can contain millions of files such as videos, digital music collections or image archives. Multimedia files are usually incompletely text-annotated, since the annotation process is burdensome. However, high-level semantic descriptions of multimedia content are very important for several application domains such as hybrid recommender and personalization systems, multimodal security applications, environmental monitoring and health monitoring systems.
The aim of this PhD thesis is to adopt state-of-the-art machine learning methodologies (such as deep networks, probabilistic topic modelling, semi-supervised learning), for content-based analysis of multimedia, by leveraging knowledge from all types of information, such as faces and facial expressions, objects, voices, music, audio events, speech emotions, user comments, web pages, video text, low level visual cues and available user/expert-generated annotations. Particular focus will be given in methods that fuse information from heterogeneous and diverse signal representations adopting deep learning techniques, as well as contextual information.

Τίτλος 3ης θέσης: Translating latent features to actionable knowledge

Περιγραφή 3ης θέσης: Complex security challenges, such as natural disasters, organized crime and mass migration, often expose the inability of governments and societies to respond timely and consistently. Despite the widespread availability of vast amounts of data and advances in analytics, effective prediction and response to adverse events remains inconsistent. In part, this can be attributed to the lack of transparency and interpretability of state-of-the-art machine learning methods, such as deep neural networks. Translating powerful latent feature representations into humanly actionable, justifiable and communicable knowledge is a challenging problem that needs to be addressed before security policy makers can fully utilize today’s far-reaching big-data and AI advances. This will allow for the creation of recurrent, human-in-the-loop, neural models where learning and actions are governed by equal levels of interpretability and justifiability.
The proposed PhD thesis will extend current state-of-the-art research at the intersection of big-data and deep learning focusing on the wider security domain. It will utilize voluminous multi-dimensional image and physical data, e.g. earth-observation and weather, for defining predictive models based on deep learning. Crucially, it will define novel semantic structures for linking multimodal input data to latent features to knowledge, building on domain-specific requirements and practices, on advances in image recognition and processing and on metadata schemas and ontologies, such as W3C PROV ( https://www.w3.org/TR/prov-overview/) and PMML ( http://dmg.org) . It will result in a novel framework for augmenting deep-learning approaches with knowledge, allowing security experts to take defensible and responsible actions efficiently.

Τίτλος 4ης θέσης: Multimodal analysis of biomedical texts

Περιγραφή 4ης θέσης: Biomedical informatics is an emerging discipline that aims to develop structures and algorithms to improve communication, understanding and management of biological and medical information. Biomedical text mining is a research subfield of biomedical informatics, dealing with the analysis of biomedical text, such as publications. Potential applications of such a text mining process include among others biological semantic indexing and/or semantic Question Answering. Most of the work done so far in this area examines text in isolation, ignoring usually the non-textual parts of the publications (like tables and images), which are often very knowledge-rich. In this PhD thesis methods and algorithms for extracting, analysing and fusing knowledge from multiple modalities in biomedical publications will be studied.

Τίτλος 5ης θέσης: Computationally-Driven Contact-Free Physiological Measurements

Περιγραφή 5ης θέσης: Research on contact-free physiological measurements is an area carrying significant scientific, social, and economic impact. The University of Houston’s Computational Physiology Lab CPL (http://cpl.uh.edu) has pioneered these measurements, reporting for the first time methods to measure heart rate, breathing rate, and electrodermal activity through thermophysiological imagery of the face. In addition, CPL has developed a state of the art tissue tracking algorithm to render all these measurements valid even in the presence of facial motion. Since then, this area has received tremendous attention, which is amplified by the advent of miniature and cheaper thermal imaging sensors.
This PhD thesis aims to conceptualize, develop, and test the next generation of contact-free physiological measurement methods exploiting also the expertise of NCSR Demokritos' SKEL Lab (https://www.iit.demokritos.gr/skel/) on data management and data analytics using machine laerning. Such contact-free methods are expected to be the standard physiological measurement methods in the next decade.


1 θέση για το Dalhousie University

Τίτλος: Big data analysis for precision medicine

Περιγραφή: Large amounts of heterogeneous medical data have become available nowadays. This introduces an opportunity and a challenge, namely to bring together these disparate data sources (e.g. scientific literature, demographic, cognitive, images, genomics, EHRs of patients) and fuse the knowledge that will be extracted. Such integrated knowledge could prove particularly valuable for improving the diagnosis and treatment of various diseases. Innovative data analysis and knowledge extraction methods could even allow the design of effective personalised treatments, helping to realise the vision of precision medicine. While data are becoming ever easier to obtain, they are difficult to process using common database management tools or traditional data processing applications. The challenges include capturing, storing, integrating, searching, sharing, and analysing. This PhD thesis aims to introduce methods and algorithms to address one of the above research challenges.

© 2017 - Institute of Informatics and Telecommunications | National Centre for Scientific Research "Demokritos"