Blogs

Transversal actions 2016-2017: Scientific events

The María de Maeztu program aimed at establishing transversal projects that contribute to an increased impact of the research activities conducted at the Department, both at international and at local level.

When creating a new program, a special focus has been put in 1) launching novel initiatives that cover an existing gap and/or supporting external pioneering projects when tey exist; 2) mobilizing novel sources of funds currently underexploited (third sector, public and private sources targeting social challenges) to promote their sustainability over time when the funds from the María de Maeztu project end.

In a series of blog posts we will cover the advance across three different segments:

  • Scientific community (this post)
  • Industrial communities, open science and innovation
  • Non-scientific communities, general public and students at all pre-university levels, with a focus on those sectors less covered by general outreach actions

SCIENTIFIC COMMUNITY

Open scholarship: The approach towards Open Science has been evolving as part of the execution of the program and impacted all the research and educational actions. The post “Developing a department of open scholars” describes the status so far.

Networks: In addition to the regular scientific communities in their areas of research, where DTIC has already a consolidated presence, DTIC is committed to contribute to the advancement of the research and academic environment in Spain, an area where the department has not been traditionally very present. In this context, in addition to the participation in specific-field scientific networks such as V-Heart SN, the Spanish network to facilitate the use of virtual hearts in the daily clinical practice, it is to highlight the active participation of DTIC-UPF in the Alliance of Severo Ochoa and María de Maeztu Centers and Units (SOMMA). DTIC-UPF is part of the members driving the workpackage on exchange of knowledge, initiated around three topics of direct work of the María de Maeztu program (Open Science, Gender, Transfer). SOMMA is composed by 41 centers in Spain, but with only 4 university departments. Therefore, our department is fully committed to support this program to guarantee that the perspectives and needs from university departments are also made visible and promoted. Another opportunity to interact with other spanish centers, in this case in the humanities, Other opportunities to promote the discussion with other disciplines, have been the participation by Xavier Serra at the ERC Conference "Frontier Research and Science Diplomacy" with CompMusic as one of the selected projects to showcase, or the participation 26th Spanish Conference of Deans in the Arts and Humanities for the discussion about the currents opportunities and threats to university departments, the perspective from an ICT department about employability and mobility and the opportunities for cooperation among ICT, social sciences and humanities.

Scientific events: A specific objective of the María de Maeztu program was to enhance the circulation of international experts in the department, to contribute to the training of researchers, promote discussion around topics of wide relevance across different research groups, enhance the visibility of our work and promote collaborations.

 

The program organised its first workshop on June 2016. The Data-driven knowledge extraction workshop provided an overview on the DTIC-UPF Strategic Program on Data-driven Knowledge Extraction, including the presentation of several ongoing projects carried out at the Department. In addition, the visit was organised to overlap with the visit of the External Advisory Board, in order to obtain useful recommendations about the current approach promoted by the María de Maeztu program and its initial steps. The Workshop included two keynote talks to contribute to the definition of department-wide discussions around two key transversal topics related to data-driven research:

  • Reproducibility in computational research by Victoria Stodden (Associate Professor. Graduate School of Library and Information Science, at the University of Illinois at Urbana-Champaign). The talk served to push the Open Scholarship strategy previously mentioned and its promotion as a core aspect of the pursuit of the research conducted at DTIC.
  • Big Data: Security and Privacy by Elisa Bertino (Professor of Computer Science, Purdue University, and Research Director, Center for Information and Research in Information Assurance and Security). Increasingly, the focus on many of the research topics addressed nowadays by an ICT department as ours is not of technological nature, but linked to its impact in individuals and society. The work at the level of the department includes research on an improved understanding of the impact of artificial intelligence on human behaviour (such as the HUMAINT project led by Emilia Gómez at the Joint Research Center of the European Commision), the impact of technology in diversity and multiculturality (with multiple examples in the music or linguistic domains) or the risks associated to algorithm transparency or privacy (subject of the research in areas such as social computing). Its strategic nature over time will probably grow, and the wide combination of disciplines at the department make it a unique environment that the program aims at fostering.

For the next two years, the program decided not to organise a similar event for two main reasons:

  • At the internal level, the regular presentations (every 3 months) by the different research projects, together with other mechanisms such as web and bulletins, guaranteed that the progress of the program was adequately followed by the staff
  • At the external level, the progress of the work of the department was generating a number of significant opportunities where scientific and industrial experts discussed the scientific, industrial, social and ethical aspects of the research of the program, or novel research directions. Therefore, it was decided to focus on increasing the impact of these actions, rather than creating new fora for the same objectives. Selected examples of these opportunities include:

The Annual Conference on Neural Information Processing Systems (NIPS) is one of the largest conferences directly linked to a wide interest in our community of researchers. The edition for 2016 took place in Barcelona and brought to our town a large number of relevant researchers working in the core topics of relevance of the program. Two actions at UPF campus allowed to bring them to our facilities, share the research advances by members of the department to relevant targets of their research, and give the opportunity to younger researchers to both interact with top researchers in their field or get involved in an event from a related field but outside their typical activities:

  • Support the organisation of the European Workshop on Reinforcement Learning (EWRL 2016), led by members of the Artificial Intelligence and Machine Learning Research Group. It is an important research event of long tradition in the machine learning research community which had a very positive impact in the increase of the recognition and visibility of the research activity carried at the DTIC department, aligned with the Maria de Maeztu excellence program, since reinforcement learning is regarded as the potential key tool for constructing interactive machine learning systems that deal with large-scale data. Reinforcement learning is an active field of research which deals with the problem of sequential decision making in unknown (and often) stochastic and/or partially observable environments. Recently there has been a wealth of both impressive empirical results, as well as significant theoretical advances. Both types of advances are of significant importance and we would like to create a forum to discuss such interesting results. EWRL had, in addition to the support of the María de Maeztu program, sponsorship from Google DeepMind and the journal Artificial Intelligence.

External speakers: Emma Brunskill (Carnegie Mellon University), Mohammad Ghavamzadeh (Adobe Research and INRIA), Hector Geffner (Pompeu Fabra University), John Langford ( Microsoft Research), Alessandro Lazaric (INRIA), Sergey Levine (University of Washington and Google), Rémi Munos (Google DeepMind and INRIA), Gerhard Neumann (TU Darmstadt), Ronald Ortner (Montanuniversität Leoben), Doina Precup (McGill University), Bruno Scherrer (INRIA), Naftali Tishby (The Hebrew University)

  • Organization of the Seminar on Music Knowledge Extraction using Machine Learning. The main goal was to have an open discussion on various topics related to machine learning applied to music, but putting special emphasis on the knowledge extraction aspects of it. Organised by members of the Music Technology Group, it is worth mentioning that it also include the participation of members of other groups such as from the Natural Language Processing Group or the Image Processing Group.

External speakers included: Cédric Févotte (CNRS), Valentin Emiya (Aix-Marseille Université, CNRS), Katherine M. Kinnaird (Brown University), Tetsuro Kitahara (Nihon University), Oriol Nieto (Pandora), Sageev Oore (Saint Mary's University), Aäron Van den Oord (Google Deep Mind), Colin Raffel (Google Brain)

  • Bioimage Analysis Workshops: A large number of events were organised around the Bioimage and Signal Analysis project, in order to promote the establishment of closer links between the biomedicine and the image processing communities, targeting to increase the visibility of the actions of the researchers working on this topic at our department. The analysis of complex data sets requires expertise in state-of-the-art image processing, statistical data analysis and mathematical modelling. Within the Maria de Maetzu program, the bia@UPF provides computational support for bioimage analysis such as: image analysis techniques; software tools/plugins usage; programming languages; script/macro writing; workflow design and buildup; user-specific analysis tools/code deliverables. The activities have included short tutortials, training courses and workshops both at UPF and externally, and in cooperation with wider communities such as the Network of European Bioimage Analysts (NEUBIAS). Examples of events hosted at UPF include the NEUBIAS Workshop (DTIC-UPF, September 13-16th 2016), the Bioimage Analysis Course and Bioimage Analysis Tools Taggathon, and the ilastik Workshop (DTIC-UPF, September 12th 2016)
  • HUman behaviour and MAchine INTelligence (HUMAINT) Kickoff Workshop (UPF, Barcelona, March 5-6, 2018). The impact of machine intelligence on human behaviour was the topic of this 2 day kickoff workshop. The workshop served to gather interdisciplinary expert knowledge (with a high representation of members of the department), to focus the programme led by Emilia Gómez at the Joint Research Center over the coming three years, and to establish a global expert community that can contribute to, and be involved in, the programme of activities over the 3 years. In addition to a high representation of members of different groups from the department in the workshop, it sought the involvement of other UPF departments such as the ones in Economics and in Life Sciences:

External speakers: Joan Serrà (Telefónica), Karina Vold (University of Cambridge), Ramón López de Mántaras (IIA - CSIC), Nicole Dewandre (Advisor in the European Commission), Lorena Jaume-Palasí (AlgorithWatch), Fabien Girardin (Near Future Lab and BBDA Data & Analytics), Verónica Dahl (Simon Fraser University), Alessandro Annoni (Head of Digital Economy Unit, Leader of Artificial Intelligence Project, European Commission), Martha Larson (Radboud University), Ansgar Koene (University of Nottingham.), Heike Schweitzer (Freie Universität Berlin), Fabien Gouyon (Pandora), Blagoj Delipetrev (Digital Economy Unit, Joint Research Centre, European Commission), Luc Steels (ICREA and CEXS-UPF), Benito Arruñada (DECON-UPF), Nuria Oliver (Vodafone Research and Data-Pop Alliance)

  • European Research Music Conference (Barcelona, June 11-13 2018). The European Research Music Conference brought together a significant number of the projects funded by the European Research Council that have music as their focus. These projects come from diverse research domains and exemplify the wide variety of high-quality research related to music that is being carried out in Europe. The Conference included talks by the Principal Investigators of the projects, plus demonstrations, short talks, and concerts by other team members. It was a great showcase of the breadth of research topics and approaches that are being supported by the most prestigious funding program of the European Commission. The European Research Music Conference was supported by ERC=Science², which is a Europe-wide communications campaign of the projects funded by the ERC and designed to reach a wide public of specialists and non-specialists in Europe and afar. Around the Conference, a number of other events and initiatives were organized in order to reach the widest possible audiences.

External invited speakers: Fang Liu (University of Reading), Paulo de Assis (Orpheus Institute), Atau Tanaka (Goldsmiths, University of London), Francois Pachet (Spotify Creator Technology Research Lab), Michael Ellison (University of Bristol), Anna Alberni and Stefano Maria Cingolani (Universitat de Barcelona), Annamaria Mesaros (Tampere University of Technology), Olivier Gouvert (Centre National de la Recherche Scientifique), Balandino di Donato (Goldsmiths, University of London), Robert Lieck (Ecole Polytechnique Federale de Lausanne), Daniel Harasim (Ecole Polytechnique Federale de Lausanne), Karl Kügle (University of Oxford), Derek B Scott (University of Leeds), Álvaro Torrente (Universidad Complutense de Madrid), Pierre Alexandre Tremblay, Owen Green, and Gerard Roma (University of Huddersfield), Margarita Díaz-Andreu (Universitat de Barcelona), Elaine Chew (Queen Mary University of London), Amanda Bayley (Bath Spa University), Jordan Smith (University of Paris-Saclay & Institut de recherche et coordination acoustique/musique), Vanessa Pope (Queen Mary University of London), Juan Parra Cancino (Orpheus Institute), Lucia D'Errico (Orpheus Institute), Michael Zbyszyński (Goldsmiths, University of London), Marco Liuni (Institut de recherche et coordination acoustique/musique), Tuomas Virtanen (Tampere University of Technology), Cédric Févotte (Centre National de la Recherche Scientifique), Shihab Shamma and Sam Norman-Haignere (Ecole Normale Superieure), Martin Rohrmeier (Ecole Polytechnique Federale de Lausanne), Jean-Julien Aucouturier (Institut de recherche et coordination acoustique/musique), Anne-Madeleine Goulet (Centre National de la Recherche Scientifique), Pablo Arias (Institut de recherche et coordination acoustique/musique),  Michela Berti (École française de Rome), Marco Cavietti (École française de Rome), Élodie Oriol (Centre National de la Recherche Scientifique), Aldo Roma (École française de Rome), Michael Clarke (University of Huddersfield), Frédéric Dufeu and Keitaro Takahashi (University of Huddersfield).

  • Other scientific workshops: Several workshops have been organised in international conferences. A list can be found here. In addition to these regular activities, the reuse to target other communities has been explored, for instance, by the repetition of the tutorial Natural Language Processing for MIR made at ISMIR2016, Columbia Univ. NYC, USA, August 7th 2016) in Barcelona in January 30th 2017 to let more students and researchers at the department take part in it, as well as an opportunity to contact and invite other local researchers working in closer areas to get to know the joint work done by the Natural Language Processing Group and Music Technology Group on the project Music Meets Natural Language Processing (all the material available here)

 

Scientific talks and seminars: The department has a consolidates series of seminars . From the perpective of the implementation of the program, it has contributed by:

  • Supporting special talks of relevance to the projects such as Data-driven design of study plans by Dr. Rakesh Agrawal, president of Data Insight Laboratories
  • Foster debates of relevance to the program on occasion of special celebrations such as an open debate among researchers at DTIC around machine learning and human behaviour, on occassion of the World Science Day for Peace and Development (Nov 10th), or on language and emotions in Wikipedia by David Laniado, from the Digital Humanities group in Eurecat, on the Wikipedia Day that celebrates its creation on January 15th. A central and diferentiating aspect of the department is the wide interdisciplinarity, and the promotion of discussion and collaboration has been central in the design of the María de Maeztu program (as an aspect to evaluate in the calls for proposals, for instance). In order to enhance this mutual knowledge, informal gatherings (specially among PhD students, and driven by them) have been promoted. The successful "Science and beers" initiative initiated at the Music Technology Group, where 3-4 researchers briefly present and discuss their work with colleagues (typically not working on the same topics) has been extended to the wider department, with the first DTIC Science and Beers event taking place in June 2018 and next editions planned after its success.  
  • Organising in the context of the PhD seminars of the department training sessions on the transversal topics addressed by the program (gender, open science, data management, transfer, impact of research). In addition to the speakers already described at the previous post for open science, it has counted with a mix of speakers ranging from senior researchers such as Gabriel Silberman (director of the Barcelona Institute of Science and Technology and Chief Scientist at Cerebri AI) and Brian Whitman (formerly co-founder of The Echo Nest, Principal Scientist of Spotify) to seminars mostly driven by PhD students of the department such as the session on "How to do good experiments online / Good practices for online experiments" where the experience of different groups has been shared.
  • Working groups: Deep learning became of huge interest to a number of groups in the department. Although the normal approach of organising seminars was already followed (with talks by reputed speakers such as - Deep Learning by Alexandros Karatzoglou (Telefónica Research); The statistical foundations of learning to control by Ben Recht (University of California, Berkeley), Music and Art with Machine LearningGenerative Models of Drawing and Sound by Douglas Eck, (Google)), the capacity to organise and give response to this interest via this mechanism is not possible. Two actions were promoted: 1) hosting talks and seminars of the wider Barcelona Machine Learning Study Group (and in addition to supporting their work, promoting that our staff and students had easy acccess to their activities), that brought talks to our campus like The Good, the Bad and the Ugly in Deep Learning byJoan Bruna (UC Berkeley), Lessons Learned from Building Real-Life ML Systems by Xavier Amatriain (Quora, video-seminar) or what matters and what doesn't? The machine learning challenges in online adds by Nicolas le Roux (Criteo); 2) supporting the creation of an autonomous study group on deep learning, which on the one hand designed and implemented their specific program based on voluntary contributions of the participants and, on the other, promoted the sharing of the work at DTIC around deep learning in different research groups via open seminars (such as Deep Learning for Transition-based Natural Language Processing. Miguel BallesterosNatural Language Processing (TALN) groupDTIC-UPF; Experimenting with Musically Motivated Convolutional Neural Networks General communication. Jordi PonsMusic Technology Group, DTIC-UPF; Machine learning, predict what you want. Ricardo Baeza, CTO CTENT and Prof. DTIC-UPF; Deep learning for learning optimal controllers. Vicenç GómezArtificial Intelligence group, DTIC-UPF). The group was open to other researchers in UPF and Barcelona. In 2018 the group has continued its activities, this time led from postdoctoral researchers from the Department of Language Sciences at UPF, located at the same campus (promoting in addition closer links among both departments). A second working group on software development is planned to be created, although efforts so far have not been fruitful.


Other events: The María de Maeztu program seeks the increase of the impact of the research conducted at the department on a very broad sense, not just limited to the traditional scientific and industrial communities. At the level of the department it has included collaboration with artists as the Artist-in-residence at the MUTEK festival within the Rapid-Mix project; the hosting of the Metadecidim Lab, an open and collaborative research space which revolves around the challenges that technology raise with respect to the transformation of politic participation; or the co-organisation with FECYT of Falling Walls Lab Barcelona 2017, an opportunity to excellent academics and professionals to present their innovative ideas, research projects and social initiatives and compete to take part in the international final in Berlin. The jury was composed by Oriol Alcoba (ESADECREAPOLIS), Esther Paniagua (Freelance Science Journalist), Ángel Sánchez (i-deals), Luca Venza(IESE) and Montserrat Vendrell (Alta Life Sciences), chaired by Xavier Serra, the scientific director of the program, and the session was chaired by Guillermo Orts-Gil from FECYT. In addition to the interesting contributions from the participants and the opportunity to DTIc researchers to take part in it, it allowed to introduce the program to he members of the jury, representatives of business schools, the entrepreneurial environment and from scientific communication.