Integrative Research Seminars 

This is a seminar series where faculties present their research in a way that is accessible to colleagues, students and researchers from other areas.

Areas of the ETIC, all to be covered, include cognitive and intelligent systems, audiovisual technologies, networks and communications, and computational biology and biomedical systems.

The integrative research seminars are open to public. 

Next IRS

 

Speaker 

Title

Abstract

Biography

Video

 

         

 

Featured IRS

 

Speaker 

Title

Abstract

Biography

Video

Thursday February 23rd, 12:30h, Auditorium

Horacio Saggion Overwhelmed by Information? Summarization, Simplification and information extraction at your fingertips.

On-line news, social networks’ posts, e-mails, Wikipedia pages, .... the amount of  information available on-line is growing at unprecedented rates. In this context of information overload, automatic text summarization and information extraction are key Natural Language Processing (NLP) techniques  for  extracting the most salient information from a document in order to populate knowledge repositories or provide users with brief stand-alone document surrogates.

Besides the amount of text we have to deal with daily, there is a further problem:  certain types of texts such as news, which are produced for a general audience,  may be too difficult to read and understand by some people, because these texts have a  complex vocabulary and grammar. In this context, automatic text simplification is another NLP  technique which  aims at transforming a text into an equivalent which, using a simpler vocabulary or syntax, would be easier to read and understand.  

Over the past years, we have been working on two lines of research; on the one hand, we have investigated ways to extract and condense the essential  information from massive scientific literature and, on the other hand, we have carried out research to produce multilingual, adaptable,  text simplification solutions for digital inclusion.    In this seminar,  I will take the opportunity to present the work carried out on our research lab on these two distinctive, exciting, and relevant topics in the information society

Bio>>  

Thursday January 26th, 12:30h, Auditorium

Luca Bonatti  Ontogenesis of rational thinking Infants possess several mechanisms to solve domain-specific problems. However, little is know about infants' abilities to reason beyond these limited domains. I will present evidence that suggests that infants have a early sense of rationality. Furthermore, when they witness a scene not previously experienced, infants reason about it by applying basic logical principles. I will argue that such inferences are used to build strategies to inspect the scenes and make inferences to enrich knowledge. I will present data about the behavioral correlates of this inferential processes in infants and adults, focusing on the case of disjunctive reasoning. Finally, I will argue that any well-designed system for human-machine interaction has to embody a model of the cognitive state of a human, in her adult and infant state. Bio >>

Thursday June 16th, 12:30h, Auditorium

Hector Geffner 

"General Solvers for General Artificial Intelligence"

Artificial Intelligence (AI) is suddenly everywhere with large corporations investing heavily in AI. Progress in many aspects of AI has been remarkable indeed yet the achievements hide a crucial limitation: the systems that excel at specific skills as a result of powerful learning algorithms are not intelligent in a broad sense. Truly flexible intelligent systems must be intentional and use predictive models and goals for different purposes: for generating behavior, for interpreting the behavior of others (humans or not), for sensing and communicating, and for learning. Learning from examples and from punishments and rewards are crucial, yet human learning is mostly about how the physical and social worlds work. In this talk, I'll review some of the progress and challenges in artificial intelligence, and our efforts aimed at achieving general artificial intelligence through the formulation and development of general effective solvers.

Bio >>

 
 

 

 

Friday May 6th, 12h, Auditorium

Ruben Moreno 

"Can we decode the brain?"

One of the highest hopes of theorists of the brain is to be able to decipher what a person or animal feels, thinks or wishes by using neuronal activity. If this were possible, then we could argue that we have understood the ‘neuronal code’, the set of symbols that the brain uses to represent and transform information. We are still very far from this desideratum to become true. However, I will describe some small steps that our lab are taking to advance into that territory.;

Bio >> 

 

April 7th, 12:30, Auditorium

Miguel A. Gonzalez Ballester

"From medical images to digital patients
and robotic surgeons"

Modern medical imaging technologies make it possible to explore the human body in a multitude of ways, depicting anatomy, metabolic function and pathological processes. Innovations in medical imaging have been paralleled by the development of computer-based image processing methods to automatically detect and diagnose pathologies, quantify disease progression and assess possible risks and complications when planning a surgical intervention. Computer support for treatment planning is not limited to image processing (filtering, segmentation, registration), but also includes artificial intelligence (machine learning, decision support systems), mathematical and computational models (statistical shape models, finite element simulations) and hardware (tracking and guidance systems, mechatronic and robotic devices). All these tools can be combined to assist in diagnosis and treatment and guide the surgeon in the operating room. In this talk, we will see examples of these techniques applied to orthopaedics, heart failure, neurodegenerative diseases, cochlear implants and foetal surgery.

Bio >> 

 

March 17th, 13, Auditorium Emilia Gomez

"Musical information retrieval: Challenges, opportunities, and classical music".

Music Information Retrieval (MIR) focuses on the research and development of computational systems to help humans better make sense of “big” music data, drawing from a diverse set of disciplines, including music theory, computer science, psychology, neuroscience, library science, electrical engineering, and machine learning. In this talk, I will provide a survey of some of my research findings in MIR that illustrate the current paradigm shift: from signal to multimodal music descriptions, from lab settings to ecologically valid evaluations, and from system-oriented to user-centric music retrieval and recommendation approaches. I will make a special emphasis on our research on the PHENICX EU project, which exploits MIR techniques to change the way we experience classical music concerts. 

Bio >> 
Feb 18th, 12:30h Auditorium Antoni Ivorra

"Healing signals: Delivery of electric current for therapeutics".

Although flow of electric current through the human body must be generally perceived as dangerous, it must be noted that controlled delivery of electric currents is the basis of some diagnostic and therapeutic approaches of great impact. Clinical applications are very diverse in terms of aim and features of the delivered currents. For instance, pacemakers deliver low magnitude and low frequency current pulses to regulate the beating of the heart whereas so called electrosurgical scalpels deliver high current RF bursts to cut biological tissues. In this talk, I will overview some fundamental interactions between living organisms and electric currents and I will briefly present the research we – the Biomedical Electronics Research Group – are conducting for understanding these interactions and for making use of them to develop new methods and devices for therapeutics. In particular, I will focus on the electroporation phenomenon and how it can be engineered for cancer treatments and on neuromuscular stimulation for neuroprosthetics by means of injectable wireless devices. Bio>>
Jan 14th, 12:30h, Auditorium Gustavo Deco

"Towards a global model of brain activity: Lessons from the human connectome"

Since the mid 1990s, the intriguing dynamics of the brain at rest has been attracting a growing body of research in neuroscience. Neuroimaging studies have revealed distinct functional networks that slowly activate and deactivate, pointing to the existence of an underlying network dynamics emerging spontaneously during rest, with specific spatial, temporal and spectral characteristics. Several theoretical scenarios have been proposed and tested with the use of large-scale computational models of coupled brain areas. However, a mechanistic explanation that encompasses all the phenomena observed in the brain during rest is still to come.We provide in this talk an overview of the key findings of resting-state activity covering a range of neuroimaging modalities including fMRI and MEG. We describe how to best define and analyze anatomical and functional brain networks and how unbalancing these networks may lead to problems with mental health. Finally, we review existing large-scale models of resting-state dynamics in health and disease
 
Bio>>
Dec 10th, 2015 *12:30h*, Auditorium  Marcelo Bertalmío "Image processing for enhanced cinematography". Despite that nowadays most cameras, even low-end consumer models, record video that when displayed on any common screen looks good and appears faithful to the original scene, on closer inspection the latter is never the case: there are details, visible at the scene, that are lost in the picture; conversely, there are details that we may only see in the picture, but were not noticeable at the scene; and colors never quite match, they don’t have the same shade, or saturation or brightness as in the original location. It is due to camera and display limitations that the shooting location and the images on the screen are perceived very differently. Adding extra lights on the location helps to improve image contrast, but this isn’t usually enough, and it’s also a very time consuming and expensive process. Trying to overcome the difference in display capabilities, professional movies are color-corrected shot by shot by skilled technicians, to generate outputs for different targets (e.g. digital cinema projectors and standard definition TV), but the results leave much room for improvement.This talk will present an overview of our ongoing work in image processing algorithms for cinema aiming to allow people watching a movie on a screen to see the same details and colors as people at the shooting location can.  Bio>>
Nov 19th, 2015 13h, Auditorium Leo Wanner "Natural language processing: From a playground of some to a central technology of many."

The computational modelling and processing of natural language (in Artificial Intelligence referred to as "Natural Language Processing" and in Linguistics as "Computational Linguistics") has been a long time a rather isolated research field. The appearance of the world wide web, the increasing popularity of intelligentagents, and the availability of large volumes of (annotated) data led to a signicant change of the perceptionof the field beyond its frontiers. Nowadays, natural language processing (NLP) technologies are central tomany information communication and processing applications - including information search and extraction,human computer interaction, decision support engines, knowledge acquisition, etc.  In the first part of this talk, we will present the state-of-the-art research carried out in the TALN group in various areas of NLP, spanning over multilingual syntactic and deep parsing, text summarization, paraphrasing and simplication, text and spoken language generation, author profiling and second language learning to more practical applications such as distillation of gists from patent material and the use of visual analytics for navigation in large scale content spaces. We will show how more classical graph transduction strategies from theoretical computer science can complement advanced machine learning-based techniques to achieve the best performance. In the second part of the talk, we will discuss how NLP can complement broader information processing technologies and illustrate this complementation drawing upon our past and current collaborative initiatives, both within the DTIC and outside.

 Bio>>

Oct 22nd, 2015

13h, Auditorium

Ralph Andrzejak

"Why is our nonlinear time series analysis group not linear?"
 

Nonlinear time series analysis allows characterizing dynamical systems in which nonlinearity gives rise to a complex, seemingly irregular temporal evolution. Importantly, these nonlinear techniques can extract information from real-world experimental signals that cannot be resolved by classical linear techniques, such as spectral analysis. Nonlinear time series analysis can, for example, help to discriminate nonlinear deterministic and linear stochastic dynamics or to characterize directional interactions between dynamics. In application to signals measured from the brain, this analysis can contribute to the understanding of brain functions and dysfunctions. It can thereby advance cognitive neuroscience and neurology.

In this seminar, I will at first illustrate some fundamental concepts of nonlinear time series analysis. Subsequently, we will review some basic facts about the disease epilepsy, epilepsy surgery and intracranial electroencephalographic recordings (EEG) from epilepsy patients. I will then show an application of nonlinear time series analysis to EEG recordings from the seizure-free interval of epilepsy patients undergoing pre-surgical diagnostics.  Results of this application allowed us to localize the seizure generating brain area without the necessity of observing actual seizure activity – a finding of clear clinical relevance. 

  Bio>>
July 10th, 2015 12:30h Auditorium Dr. Josep Blat "Some GTI Research: HCI in the Wild and Democratisation of 3D Graphics" Some of the ongoing research in HCI and CG, 2 of the 3 areas GTI works on (the third being Technology Enhanced Learning), will be discussed, framed in ongoing research trends, the so called "HCI in the wild", and democratisation of 3D graphics, respectively   Bio>>
 June 4th , 2015  Dr. Xavier Serra

"Research highlights in my journey  within the field of Sound and Music Computing"

In this presentation I will go over some of the research I have been involved with in my thirty-year career within the field of Sound and Music Computing, emphasizing the goals I aimed for and identifying some of the results obtained. 

My personal research, and the one I have directly supervised, has been mainly focused on the analysis, description and synthesis of sound and music signals. My initial efforts were dedicated to analyze and transform complex and musically relevant sounds; sounds that were not well captured by the audio processing techniques used at that time. My approach was to use spectral analysis and synthesis techniques to develop a deterministic plus stochastic model with which to obtain sonically and musically meaningful parameterizations and descriptions. That work had practical applications for synthesizing and transforming a wide variety of sounds, including the voice.

My research has always been motivated by music, by the interest of developing musical tools that can be socially and culturally relevant. In this talk I want to emphasize this aspect while talking about my thirty-year research journey.

 Bio>>
 April 10th 2015. 13:00h – 14:00h. 
Auditorium 
 Bart Bijnens  "Where engineering meets clinical practice: understanding the working  of the heart in health and disease" Contemporary research in cardiovascular medicine is aimed at both the understanding of the fundamental properties of the cardiovascular system and its changes with disease, as well as providing evidence-based guidelines towards best clinical practice. Linking basic understanding with clinical practice is the objective of translational biomedicine, in which "reality-driven" research is key, underlining the concept that direct human observations are essential for the study of hypotheses relevant to human reality [Mankoff S, Journal of Translational Medicine 2004]. However, reality-driven biomedical research is accompanied by important hurdles since interpreting complex observations and linking these to relevant basic and clinical understanding of the human and their diseases require an interaction of basic scientists from all disciplines, technologists and computer scientists and clinical specialists, all speaking different languages.   Bio>>