Back

Visit by the Basque Center for Applied Mathematics - BCAM

We are pleased to have a group of researchers from the Basque Center for Applied Mathematics (BCAM) visiting our Department, around their participation in the 7th PhD Workshop of the Department. We look forward to this visit, and we hope it becomes the seed for many fruitful future cooperations. Especifically, the visitors will be José Antonio Lozano, the director of the center and leader of the machine learning research line at BCAM; Santiago Mazuelas, Ramón y Cajal researcher at the macine learning line, and five PhD students: Amaia Abanda Elustondo (machine learning), Iker Beñaran (machine learning), Tamara Dancheva (CFD Computational Technology), Sandeep Kumar (Linear and Non-Linear waves) and Isabella Marinelli (Mathmatical Modelling in Biosciences). They will take part in the activities described below, but if you would like to have additional meetings, feel free to contact them or Aurelio Ruiz to arrange it.

The day before the workshop Santiago Mazuelas will give a seminar (15:30h, room 55.309, webpage for seminars) titled "Data processing under uncertainty and tractability limitations” (see full details below).

On April 4th, the following 5 PhD students from BCAM will join the PhD Workshop (see abstracts below):

  • Amaia Abanda Elustondo. Discriminatory Features Extraction for Algorithm-Type Selection in Time Series Classification
  • Iker Beñaran. Crowd Learning with Candidate Labeling: an EM-based Solution
  • Tamara Dancheva. Distributed memory algorithms for Finite Element computing in FEniCS with applications in continuum mechanics
  • Sandeep Kumar. On the evolution of Vortex Filament Equation for regular M-polygons with nonzero torsion
  • Isabella Marinelli. On the Integrated Oscillator Model for Pancreatic β-cells: Development, Results, and Applications

In addition, on April 8th DTIC-UPF and BCAM, as leaders of the working group on Open Science in the network of Severo Ochoa and María de Maeztu Centers and Units of Excellence (SOMMA), organise an Open Science Day.

DETAILS

Seminar:

Santiago Mazuelas, April 3rd. 15:30h, room 55.309. "Data processing under uncertainty and tractability limitations” (see full details below).

General data-driven problems can be seen as decision problems in which actions are chosen with the aid of data. For instance, supervised classification uses training data to choose a classification rule; and sequential inference uses time series to sequentially choose a target variable. 

The development of data processing techniques to address such problems is hindered by the fact that actual probability distributions of state variables are often unknown or intractable in practice. For instance, in supervised classification the actual distribution of features-label pairs is often unknown, and in sequential inference the actual posterior distribution of target values is often intractable. 
In this talk, I will describe approximation techniques to address uncertainty and tractability limitations based on optimization methods. In particular, I will present data processing techniques that address distributions' uncertainty in supervised classification and distributions' intractability in sequential inference.
 

Santiago Mazuelas received the Ph.D. in Mathematics and Ph.D. in Telecommunications Engineering from the University of Valladolid, Spain, in 2009 and 2011, respectively. Since 2017 he has been Ramon y Cajal Researcher at the Basque Center for Applied Mathematics (BCAM). Prior to joining BCAM, he was a Staff Engineer at Qualcomm Corporate Research and Development from 2014 to 2017. He previously worked from 2009 to 2014 as Postdoctoral Fellow and Associate in the Wireless Information and Network Sciences Laboratory at the Massachusetts Institute of Technology (MIT). He is a frequent visitor of the Laboratory for Information and Decision Systems (LIDS) at the MIT, where he holds the Research Affiliate appointment. His general research interest is the application of mathematics to solve engineering problems, currently his work is primarily focused on statistical signal processing, machine learning, and data science. Dr. Mazuelas is Associate Editor for the IEEE Communications Letters and served as Co-chair for the Symposiums on Wireless Communications at the 2014 IEEE Globecom and at the 2015 IEEE ICC. He has the Young Scientist Prize from the Union Radio-Scientifique Internationale (URSI) Symposium in 2007, and the Early Achievement Award from the IEEE ComSoc in 2018. His papers received the IEEE Communications Society Fred W. Ellersick Prize in 2012, and Best Paper Awards from the IEEE ICC in 2013, the IEEE ICUWB in 2011, and the IEEE Globecom in 2011. 

Posters presented at the PhD Workshop (see abstracts below, full program of the PhD workshop here)

  • Amaia Abanda Elustondo. Discriminatory Features Extraction for Algorithm-Type Selection in Time Series Classification

Time series classification (TSC) is a particular case of classification in which the input instances are ordered sequences. Due to their special nature, the research community has assumed that specific methods are required for dealing with this type of data. These specific methods have been categorized in the literature as elastic, interval-based, shapelet-based, and dictionary-based. Elastic classifiers are able to handle time warping and shifting, interval-based classifiers deal with phase dependent discriminatory subseries, shapelets-based classifiers with time independent discriminatory patterns and dictionary-based classifiers take into account the frequency of these patterns. One of the main problems is that, in practice, when a new dataset is available, it is not clear which of the different types of algorithms for TSC will work better. The aim of this work is to extract features from time series datasets that inform about which type of algorithm is more suitable in each case.

  • Iker Beñaran. Crowd Learning with Candidate Labeling: an EM-based Solution

Crowdsourcing is widely used nowadays in machine learning for data labeling. Although in the traditional case annotators are asked to provide a single label for each instance, novel approaches allow annotators, in case of doubt, to choose a subset of labels as a way to extract more information from them. In both the traditional and these novel approaches, the reliability of the labelers can be modeled based on the collections of labels that they provide. In this paper, we propose an Expectation-Maximization-based method for crowdsourced data with candidate sets. Iteratively the likelihood of the parameters that model the reliability of the labelers is maximized, while the ground truth is estimated. The experimental results suggest that the proposed method performs better than the baseline aggregation schemes in terms of estimated accuracy.

  • Tamara Dancheva. Distributed memory algorithms for Finite Element computing in FEniCS with applications in continuum mechanics

FEniCS is an open-source computing framework for automatic resolution of general partial differential equations. FEniCS-HPC is a highly efficient framework that has been based on FEniCS with focus on performance and scalable distributed memory algorithms suitable for supercomputer architectures. The application includes problems in computational medicine, renewable energy technology, aerodynamics and various other computational fluid dynamics and computational solid mechanics problems. The ENABLE project is a European Training Network whose main goal is the development of new solutions for the simulation of various manufacturing processes of metal alloys using the FEniCS and FEniCS-HPC platforms. One key development has been a promising approach based on the Omega_h library for general mesh modification with distributed memory. The main motivation behind the mesh adaptation is to prevent degradation of elements and limit the waste of resources in large strain, deforming/moving mesh simulations such as the ones on which the ENABLE project focuses on. We apply this approach to different problems: fluid-structure interface, aerodynamics and fracture simulations.

  • Sandeep Kumar. On the evolution of Vortex Filament Equation for regular M-polygons with nonzero torsion

The Vortex Filament equation describes the self-induced motion of a vortex filament in three dimensions and some of its explicit solutions are a straight line, circle and, helix. In their recent work on regular planar M-polygons, de la Hoz and Vega conjuncture that the dynamics of a circle is not stable. This fact is proved with both theoretical and numerical arguments by approximating the circle with M sided polygons. In this poster, we show that by introducing a nonzero torsion a similar approach would extend the results for helix and straight line. The trajectory of a single point which exhibits a multi-fractal phenomenon and resembles to the Riemann's non-differentiable function(RNDF) for the planar polygons, remains to be multi-fractal for the non-zero torsion case also and is related to a version of RNDF. We provide appropriate algebraic results to support our numerical observations and to describe the evolution of polygonal curves.

  • Isabella Marinelli. On the Integrated Oscillator Model for Pancreatic β-cells: Development, Results, and Applications

ATP-sensitive K+ channels play a fundamental role in the insulin secretion mechanism of pancreatic β-cells by coupling metabolism to cellular electrical activity. Genetic mutations affecting these channels’ function can result in neonatal diabetes, currently treated with high- risk drugs. To improve treatment, an exhaustive physiological under- standing of the β-cell function is needed. The Integrated Oscillator Model (IOM) is used to investigate the mechanism behind bursting activity that underlies intracellular Ca2+ oscillations and pulsatile in- sulin secretion. In this model, cellular electrical activity, intracellular Ca2+, and glucose metabolism interact via numerous feedforward and feedback pathways. These interactions produce metabolic oscillations with distinct time courses, reflecting different oscillation mechanisms. After identifying conditions favorable to each type of oscillations, we show that IOM is able to reproduce key experimental findings of β- cell activity. Extension of the model can describe complex biological processes like the insulin exocytosis cascade or the interaction among β-cells within the islet.