Research Lines

The world is increasingly driven by software, being a key part of most disruptive innovations such as autonomous vehicles, virtual reality or 3D printing. However, poor software applications can have dramatic consequences in term of usability, security, money, and even people safety. It is therefore crucial to have the right tools for the development of high-quality software systems that meet user requirements and budget constraints. Software engineering research at SCORE focuses on the development of techniques and tools for the development of more efficient, effective, and reliable software applications.

Some of our main activities in SCORE are related to the development of software services. Specifically, researchers at SCORE work on new languages for their specification and on the development of tools for automating their optimal deployment and operation using techniques like constraint programming and metaheuristics.

Another important area of expertise in SCORE is in the automated detection of faults in software applications. In recent years, SCORE researchers have played a leading role tackling the test oracle problem, recognized as one of key challenges in software engineering. Their advances in the field of metamorphic testing have led to the detection of bugs in software services with millions of users world-wide. Currently, researchers at SCORE explore the intersection between software testing and artificial intelligence to automate much of the testing tasks currently perform by humans.

Search-based software engineering is another relevant research topic in SCORE. The goal is to tackle relevant software engineering challenges by translating them to optimization problems and proposing specific solutions using techniques like evolutionary algorithms and artificial intelligence. Among others, researchers at SCORE have proposed relevant search-based solutions for the problems of optimal service composition, test case selection and prioritization, and optimal product selection in product lines.

The research line on Software and Service Engineering is led by Sergio Segura (0000-0001-8816-6213), and it is composed of 8 members of the research staff: 2 Catedráticos de Universidad, 5 Titulares de Universidad, and 1 Contratado Doctor.

Natural Computing is a discipline inspired by the operation of living organisms. It has at its core the simulation and implementation of the dynamic processes that occur in nature and which are susceptible to interpretation as computing procedures.

As such it offers an alternative to conventional means of computation, contributing new paradigms that can provide effective solutions to some of the limitations of current computing.

Well-know classical examples are Genetic algorithms (Holland 1975) and Artificial neural networks (McCulloch and Pitts 1943), but SCORE researchers also have experience in DNA computing (Head 1987, Adleman 1994) and specially in Membrane computing (Paun 1998).

SCORE research activities within this line focus on the interplay between Computer Science, Mathematics, Biology and Population Dynamics. Specifically, the main goal is the development of enabling technologies based on bio-inspired formal methods.

More precisely, we combine three different approaches:

1. Theoretical foundations. Formal definition of unconventional models of computation, and analysis of their computational power and efficiency (e.g. tackling the P vs NP problem).

2. Software implementation. Formal specification of the semantics of these models of computation, and development of the corresponding software (parsers and simulation engines). Development of high-performance simulation tools using massively parallel architectures, mainly GPU computing. Development of flexible Virtual Research Environments, easily allowing customization of the GUI, the expected input parameters, and the output data.

3. Applications.

Wide range of applications to various engineering areas, including engineering optimization, power system fault diagnosis, social robotics, controller design of mobile robots, or complex systems involving data modeling and process interactions.

Applications to biological case studies, ranging from molecular interactions (e.g. signalling pathways, or inter-bacterial communication) to macroscopic systems (e.g. population dynamics of endangered/invasive species, or virus spread modelling).

This research line is led by Agustín Riscos-Núñez (0000-0002-5409-3578) and it is composed of 4 members of its research staff: 1 Emmeritus Professor and 3 Titulares de Universidad.

Neuromorphic computing copies observed behaviors of biological neural networks through analogue or digital electronic circuits. The main aim is twofold: applying brain/neural inspiration to generic cognitive computing to solve engineering problems and offering a tool for neuroscience to understand the dynamic processes of neural systems, like learning and development in the brain.

The concept of neuromorphic computing was pioneered by Caltech professor Carver Mead in the 1980s. But neuromorphic computing (sometimes called neuromorphic engineering) is still considered an emerging field, and only in the last few years has it become feasible for commercial use cases. To mimic the human brain and nervous system, researchers are building artificial neural networks that replace synapses with nodes. One of the obstacles to these networks is the binary nature of digital processing. The use of CPUs (based on Von-Neuman architectures) implies to send messages through circuits that are either on or off. But neurons are more complex, and they need to work with deeper range of values on their membranes, which evolves according to the synapse dynamic behavior, and send spikes (on or off) to other neurons depending on properties that relay on the dynamics of those membranes.

The engineering of a neuromorphic device involves the development of components whose functions are analogous to parts of the brain, or at least to what such parts are believed to do. To this end, engineers require new tools able to manage precisely the timing response of the neurons and respect their model. Analog circuits are clear candidates to satisfy these requirements. In the other hand, new computing architectures in the digital domain are raising for the improvement of this field.

As a result, neuromorphic engineers are currently building different approaches:

• Mixed analog-digital processors that can modulate the amount of current flowing between nodes, like the varying strength of electric impulses in the brain that form and alter brain chemistry. Examples are Dynap-SE from SynSense Ltd. or IBM TrueNorth chip.
• Digital architectures based on classic CPUs arranged for improving the timing considerations of biological neural networks for its right modeling, like SpiNNaker, based on ARM processors from University of Manchester, or Loihi from Intel.
• Sensors that mimics the behavior of our senses offering a spike-based representeatiion of the information, like the dynamic vision sensors (ie. DAVIS from University of Zurich), analog cochlea (ie. AEREAR from Institute of Neuroinformatics) or digital models (ie. Neuromorphic Auditory Sensor from University of Seville).
• FPGA circuits that mimic the spiking sensors information in similar ways as it is filtered and adapted in the first layers of our brains (ie. Spinking Convolutional Layers from University of Sevilla and IMSE-CSIC).
• Digital implementations of spike-based control algorithms for motor-neuron emulation and its application to robotics (ie. SPID and insect inspired SCPG from University of Sevilla and University of Cadiz).

Neuromorphic engineers seek to solve current computing paradigms in an improved way taking inspiration from the brain, which is more efficient in many computational tasks and much more efficient in terms on energy consumption. This operation is focused on uncertainty, that means to deal with probabilities instead of deterministic models (i.e. using analog computation or quantum computing). And looking for solutions that bring the memory closer to the computation itself (i.e. exploring memristors). With these ingredients, neuromorphic computing is currently focused on evolve the machine learning to allow computer to learn through observation and practice, rather than programming or training through vast number of samples from a dataset.

Currently, neuromorphic processors can simulate up to 16 billion synapses —still far from the brain’s 800 trillion. This type of processor remains an infinitesimally small piece of the market and is mostly intended for research and defense purposes. But the technology has proven theoretically feasible, and if you believe some experts, practical and commercial applications are only a matter of time.

Key advantages of neuromorphic computing compared to traditional approaches are energy efficiency, very reduced latency, real-time operation, robustness against local failures and the ability to learn.

This research line members have been working in Neuromorphic Engineering for 20 years and have participated several times in key workshops, like Telluride Neuromorphic Cognition Engineering Workshop (USA) or The CapoCaccia Workshops toward Neuromorphic Intelligence (EU). It is focused on neuromorphic digital circuits developments (using both FPGAs, ASICs and embedded systems), neuromorphic auditory sensors, neuromorphic motor-control for robotics, spike-based information processing, deep-learning using neuromorphic sensors, and spike-based deep-learning algorithms implementations.

The research line on Neuromorphic Systems is led by Alejandro Linares-Barranco (0000-0002-6056-740X) and it is composed of 3 members of the research staff and 10 collaborators.

Information Systems play a key role in organizations nowadays. As a discipline, Information Systems cover a wide range of research topics, from the study of the impact on the use of Information Systems in organizations to the design and development of new technologies that enable new ways of work. At SCORE, our research on Information Systems focuses on the development and application of software tools to improve the performance and human resource management of business processes with a particular emphasis on unstructured knowledge-intensive processes.

In the field of performance management, the group has a strong experience in the monitoring of business processes based on process performance indicators (PPIs). The current interests involve improving the modelling, monitoring and prediction of PPIs. Regarding modelling, the research is focused on making the definition of PPIs and the whole process to develop a PPI dashboard easier for non-expert users. Concerning monitoring, new techniques and methodologies for the definition and monitoring of decisions and unstructured processes are being devised. Finally, in the area of predictive monitoring of PPIs, the research targets problems that appear when a predictive model is deployed in a production system, such as the reliability of the models or the evolution of the predictive model.

The research on human resources covers several different angles. One stream of research focuses on the application of methodologies to improve personal productivity. This includes the analysis of the effect of techniques like mindfulness to perform cognitive-intensive tasks like conceptual modelling, and the development of novel methodologies for time management and work organization. Another research stream is focused on the configuration and use of workstream collaboration tools and other related technologies to improve the collaboration of people in a context of digital transformation. Finally, the third research stream tackles the organizational perspective of business processes pursuing the optimization of the management of human resources along with process modelling, execution and analysis.

This research line is led by Manuel Resinas (0000-0003-1575-406X) and it is composed of 5 members of its research staff: 2 Titulares de Universidad, 1 Contratado Doctor, 1 Ayudante Doctor and 1 PI of a JIN project.