loading...

Doctoral Program

Doctoral Program

Research Topics for PhD Theses

This page contains a list of possible research themes for the PhD Thesis, and respective supervisors. Prospective students interested in pursuing research in any of these themes should contact the supervisors to obtain additional information. When applying to the Doctoral Program, students interested in a specific research theme (possibly, but not necessarily, from this list), should state their intention in the Motivation Letter.

Remarks:
  • This list of research themes is not exhaustive, either regarding themes and potential supervisors; prospective students are invited to consult information about the research groups of the research center that supports this Doctoral Program (CISUC).
  • This list is not closed: new proposals are likely to be included in the page at any moment, so we recommend prospective students to return regularly.

Research Topics

Title Supervisor(s)
Development of novel features and algorithms for epileptic seizures prediction César Teixeira (University of Coimbra, PT)
Felipe França (UFRJ, BR)
Adaptable Computer Models for Interactive Sound Arts Jose Fornari (UNICAMP, BR)
Amílcar Cardoso (University of Coimbra, PT)
Live Interactive Composition, Evolutionary Computation and Models derived from Computational Neuroscience Jonatas Manzolli (UNICAMP, BR)
Amílcar Cardoso (University of Coimbra, PT)
Towards Multimodal Processes for Artistic Production Artemis Moroni (UNICAMP, BR)
Penousal Machado (University of Coimbra, PT)
Relevance Feedback for Semantic and Aesthetic Retrieval of Images Ricardo Torres (UNICAMP, BR)
Penousal Machado (University of Coimbra, PT)
Sequence Alignment Methods for the Analysis of Robustness Testing Results Eliane Martins (UNICAMP, BR)
Luís Paquete (University of Coimbra, PT)
Methodology for Agile Development integrated with Acceptance Testing Eliane Martins (UNICAMP, BR)
Marco Vieira (University of Coimbra, PT)
Cecília Rubira (UNICAMP, BR)
An Adaptive Approach for Fault-Tolerant Service Applications Cecília Rubira (UNICAMP, BR)
Marco Vieira (University of Coimbra, PT)
Eliane Martins (UNICAMP, BR)
Assessing and Improving the Trustworthiness of Service-Oriented Architectures at Runtime Marco Vieira (University of Coimbra, PT)
Eliane Martins (UNICAMP, BR)
Cecília Rubira (UNICAMP, BR)

Details

Title

Development of novel features and algorithms for epileptic seizures prediction

Supervisor(s)

César Teixeira (University of Coimbra, PT), Felipe França (UFRJ, BR)

Description

Among the individuals that suffer from epilepsy, some are resistant to the available medication and does not include the conditions required for surgery. These patients will benefit from devices running algorithms that forecast and warn the upcoming of a seizure. Features extracted from the EEG, in time and frequency domains, have been used in the research of efficient algorithms for seizure prediction. Although significant progresses have been made, namely in the European Project EPILESIAE coordinated by CISUC, there remains still the need for further developments of computational intelligence and machine learning algorithms to decrease the false prediction rate, in order to give clinical validity to the developed alarming devices.

This thesis proposal will be focused on the research of new features, and new classification strategies, to approach seizure prediction to levels appropriate for real-world application.

Multivariate features, using the available multichannel EEG signals, may have the potential to capture the dynamic behaviour of the brain in a more effective way than the single-channel (univariate) features used so far. In this thesis, new features and improved combination of univariate and multivariate features will be investigated.

The application and potential new developments of specialized neural networks shall be also investigated, such as recurrent networks and weightless neural networks, in order to accommodate dynamic changes and to improve the generalization capability to long-term EEG data.

The thesis will be based on the EPILESIAE database that contains data from 278 patients. For each patient it is available long-term (approx. 5 days), continuous and annotated recordings.

The prospective candidate should have bases of analogue and digital signal processing, as well as of computational intelligence methods (artificial neural networks, support vector machines, etc.). Advanced programming skills on Matlab and C will be also required.

top

Title

Adaptable Computer Models for Interactive Sound Arts

Supervisor(s)

Jose Fornari (UNICAMP, BR), Amílcar Cardoso (University of Coimbra, PT)

Description

This project aims to study, develop and implement adaptable computer models such as generative and evolutionary ones for their usage in sonic arts performances. These models are implemented in Pd (www.puredata.info), an open-source computing environment for the development of real time data analysis, transformation and synthesis. This aims to touch, in artistic terms, one of the foremost philosophical problems, which is the understanding of how someone, such as an artist, interacts with the external reality, while internally creating his/her artwork. How does the reality shapes and intervenes the artist creative process. For that we take the assumption that human mind understands, recognizes and rapport with reality through a constant and dynamic process of mental-modeling. The process is here seen as divided in three states: 1) Perception, where the mind receives sensory information from outside, throughout its bodily senses, which comes from distinct mediums, such as mechanical (e.g. hearing and touch), chemical (e.g. olfaction and taste) and electromagnetic (e.g. vision). According to evolutionary premises, these stimuli are non-linearly translated into electrochemical information to the nervous system. 2) Cognition; the state that creates, stores and compares models with the information gathered or from previously reasoned models. This is the information processing stage. 3) Affection, where emotions arouse, as an evolutionary strategy to motivate the individual to act, to be placed “in-motion”, in order to ratify, refute or redefine the cognitive modeling of a perceived phenomenon. This project aims to use adaptable computer models that (to an extent) emulate such perceptual, cognitive and affective mental processes into algorithms (Pd patches) for the interactive exploration of analysis, composition and performance of music as well as other types of sonic arts, such as soundscape design.

top

Title

Live Interactive Composition, Evolutionary Computation and Models derived from Computational Neuroscience

Supervisor(s)

Jonatas Manzolli (UNICAMP, BR), Amílcar Cardoso (University of Coimbra, PT)

Description

This project intends to explore the convergence between ideas developed around contemporary music and ideas that have emerged from models in evolutionary computation and computational neuroscience. Music composition has evolved from symbolic notated pitches to expressions of the internal organization of sound. This can be observed in the extended instrumental techniques developed from the 1940’s onwards up to the more recent compositional strategies that have emerged from the “new interfaces for musical expression”. The dynamic organization of sound material in “real” time, however, adds new dimensions to musical information and to its symbolic representations. We are studying a conceptual point of view according to which a theory of mind can be applied to the development of systems that produce sound material in time, through the interaction of simple rules and independently of symbolic notation and the application of evolutionary computation to direct the internal dynamics of such systems. Conversely the real-world realization of such artifacts might also be considered as the only way to further develop such a theories of mind. In (Verschure & Manzolli, 2013) we describe this research approach illustrating our point of view and present a number of musical systems that have been realized as experiments of such a “situated aesthetics” based on cognitive models of interaction.

top

Title

Towards Multimodal Processes for Artistic Production

Supervisor(s)

Artemis Moroni (UNICAMP, BR), Penousal Machado (University of Coimbra, PT)

Description

This research is oriented towards the study of automatic and semi-automatic processes of artistic production, exploring the arbitrariness. The concept arose from the attempt of computationally emulate creativity applied to artistic production in the visual and sound domains by using evolutionary computation, fuzzy logics and neural networks. An emergent question is: are there rules that guide aesthetical appreciation? If aesthetical appreciation would be governed only by subjective opinion, it would not be possible to obtain (partially) automatic shapes of artistic production, with some aesthetical value, without a complete integration of the user with the machine. On the other hand, if general rules did not allow the maintenance of a set of degrees of freedom of expression, there could be complete automation, in spite of the possible complexity of design. From this perspective, human and machines are agents of a complex system and the artifacts emerge from their interaction and behavior. Different setups explore levels of autonomy between humans and machines. The creative process of the composer, artist, engineer, the audience and performers, including robots, their way of working and the mechanisms of the different areas involved are examined with results where the material and the idea are blended in the final result. There is interest in expanding this research to the area of creative text generation.

top

Title

Relevance Feedback for Semantic and Aesthetic Retrieval of Images

Supervisor(s)

Ricardo Torres (UNICAMP, BR), Penousal Machado (University of Coimbra, PT)

Description

Technology advancements towards image acquisition, storage, and dissemination have fostered the creation of large image collections. In this scenario, there is a strong demand of solutions to perform effective an efficient image searches. One of the widely used approaches relies on the use of content-based image retrieval (CBIR) systems. A CBIR system aims to define relevant images in the collection, by taking into account their similarity to a given query pattern (e.g., query image) usually measured in terms of their visual content properties (such as color, texture, and shape). However, for complex queries over heterogeneous collections, low-level visual features are not able to properly describe the rich semantic intent of a query, nor the high-level concepts found within images, a problem known as semantic gap. Another issue is that different users quite have different interpretations of an image and different aesthetic preferences. Moreover, the interpretation and preferences of a given user may change over time.

In order to make systems adaptable to different users, relevance feedback approaches have been proposed as a suitable alternative. In these interactive search systems, users can implicitly refine retrieval results by explicitly providing relevance feedback for the items in the result list. The system can use this information to expand queries and improve internal learning models, providing new results that are supposed to better fulfill the user’s needs. Several studies have shown the capacity of relevance feedback systems for improving the retrieval effectiveness when only semantic aspects are considered. In these systems, aesthetic preferences of users are not considered in the retrieval process. Thus, although the semantics of the retrieved images tends to match the user’s intents, their aesthetic qualities are often inadequate to fulfill the user’s needs. In this project, we aim to design and implement relevance feedback systems that combine image representations and learning mechanisms that exploit both image semantics and aesthetic criteria.

top

Title

Sequence Alignment Methods for the Analysis of Robustness Testing Results

Supervisor(s)

Eliane Martins (UNICAMP, BR), Luís Paquete (University of Coimbra, PT)

Description

For the analysis of robustness testing results, comparison with a golden run is commonly used to determine whether robustness failures occurred or not. One limitation of this approach is that traditional comparison techniques require the system to have a repeatable behavior under the same workload, whether or not in presence of faults. This limits the applicability of this technique, since most systems have indeterministic behavior due to concurrency, for example. Moreover, tools that use golden-run comparison hardly ever give information about what are the common patterns of behavior between the golden-run and a faulty trace and where there are deviations due to action of fault-tolerance mechanisms, for example. The goal of this project is to use sequence alignment algorithms for golden run comparison. One of the challenges in using sequence alignment is the determination of a scoring scheme that leads to a meaningful alignment to the system under test. This is difficult even in Biology, for which there are historical data about evolution to guide the construction of such schemes. In particular, we consider the multi-objective variant of the sequence alignment with the aim of helping in the determination of such a scoring scheme, which brings the advantage of providing a set of alignments that represent the trade-off between performing gaps and finding matching symbols from the sequences which are meaningful for the systems under analysis,  and, therefore, bringing more information to the user. This is particular interesting in the context of search-based software engineering, which lacks the use of multiobjective methods for test results analysis, as opposed to other application areas.

Of particular interest to this project is: i) the automatic generation of appropriate substitution score matrices according to the semantic of the system being tested; ii) to develop heuristic and exact methods for the multi-objective variant when more than two sequences are considered;  iii) to derive automatic methods for selecting the most interesting alignment from the trade-off set. The methods will be validated on a wide range of simulated and real-life problems.

top

Title

Methodology for Agile Development integrated with Acceptance Testing

Supervisor(s)

Eliane Martins (UNICAMP, BR), Marco Vieira (University of Coimbra, PT), Cecília Rubira (UNICAMP, BR)

Description

Acceptance Testing can be used in Agile Methods to guide the development as well as the testing activities. However, some open questions still remain, such as: what to test? What not to test? How much to test? Besides, exceptional behavior generally is not specified, and hence, not adequately tested.

This project aims to provide a method to guide the integration of model-based testing to Agile Software Development, in order to make use of the advantages of model-based testing while also keeping the necessary agility in the system’s development. The method will also integrate Scrum with treatment of Exceptional Behavior, allowing them to support the creation of more detailed tests. To accomplish this, Exceptional User Stories and high-level architecture artifacts will be used, which will also support the development and the testing activities. Exposing exception handling at early development stages aims to achieve high reliability of the system.

As Behavior Driven Development (BDD) proposes, Acceptance Tests are derived from User Stories and represented by Scenarios. In this project, behavior models will be derived from User Stories and Exceptional User Stories to support the scenarios definitions. Based on these behavior models and using model-based testing techniques, robust scenarios will be derived. The behavior models will be constructed in parallel to the construction of the architecture artifacts, so that both will be used by developers to build the systems and also by testers to verify and validate the system. All the artifacts used in this project have to be flexible and simple in order to better fit Agile Software Development. Also, they will be constructed incrementally.

top

Title

An Adaptive Approach for Fault-Tolerant Service Applications

Supervisor(s)

Cecília Rubira (UNICAMP, BR), Marco Vieira (University of Coimbra, PT), Eliane Martins (UNICAMP, BR)

Description

Today’s society is highly dependent on systems based on Service-Oriented Architectures (SOA) for its basic day-to-day functioning. A composite service, the basis for the construction of applications in the SOA world, can be regarded as a combination of activities invoked in a predefined order and executed as a whole. Nevertheless, it is unlikely that services and service compositions, usually controlled by third parties, will ever be completely free from software faults arising out of wrong specifications and incorrect coding.

SOA-based systems often rely in an environment that is highly dynamic and several decisions should be postponed until runtime, where different stakeholders with conflicting requirements exist, and fluctuations in the quality of services (QoS) are recurrent. Most of the service-based applications proposed in the literature do not support an adaptable architecture in which changes can be made at runtime. Ideally fault tolerance mechanisms, in particular exception-handling mechanisms, for SOAs should adapt themselves to bring out the most appropriate strategy for error handling in close accordance with the clients’ requirements and the environment.

Dynamic Software Product Lines (DSPLs) provide a modeling framework to understand a self-adaptive system by highlighting the relationships among its components. DSPLs extend Software Product Lines (SPLs) to support late variability. Software variability is defined as the capacity that a software system or a software artifact has to be modified for use in a particular context at some point in its lifecycle. At the design phase, variability can be structured into Product Line Architectures (PLAs) by means of variation points. DPLSs allow a dynamic binding of variants in order to satisfy a variation point.

The goal of this PhD work is to propose a DSPL infrastructure to support a family of fault-tolerant service-based applications so that the most suitable error handling strategy can be instantiated at runtime in response to changes in the clients’ requirements and in current context.

top

Title

Assessing and Improving the Trustworthiness of Service-Oriented Architectures at Runtime

Supervisor(s)

Marco Vieira (University of Coimbra, PT), Eliane Martins (UNICAMP, BR), Cecília Rubira (UNICAMP, BR)

Description

Service Oriented Architectures (SOAs) are nowadays used in a wide range of organizations and scenarios, including business-critical systems. These architectures consist of several interacting software resources (services) that are designed to support the information infrastructure of an organization. These architectures present particular characteristics as high complexity, extreme dynamicity, and a very large scale of composable components/elements and services. The forthcoming evolution is expected to exacerbate this trend even more, together with other evident facets, such as the needs for high mobility, scalability, and flexibility.

Complying with nowadays organizations’ requirements demands for the deployment and maintenance of trustworthy dynamic service-based software systems, which naturally results in the superposition of the design and runtime phases, thus imposing the need for continuously system assessment and adaptation. Assessing trustworthiness can be seen as the continuous process of quantifying and exposing the trustworthiness relationship between a system and its users, and represents a powerful alternative to traditional dependability and security assessment.

Unfortunately the detailed assessment of a system prior to its deployment does not fit a service-oriented context where a multitude of services is being deployed, interconnected and updated continuously, and where runtime composition and evolution plays a central role. To overcome this problem new runtime assessment and improvement approaches are necessary, assuring the required quality of dynamic and evolving service oriented architectures.

The goal of this PhD project is to define approaches for continuous trustworthiness evaluation and improvement. In practice, we will research techniques that take advantage of monitoring services and infrastructures to support the runtime assessment of the system through the collection of measurements for quantitative analysis of trustworthiness. Such assessment should serve as input for the implementation of continuous architectural adaptation techniques that lead to the improvement of the system trustworthiness.

top