The present proposal for Europa Investigación 2019 aims to consolidate an existing Innovative Training Networks (ITN) project proposal related to Artificial Intelligence (AI) for the upcoming calls of H2020-MSCA-ITN-ETN actions for years 2020 and 2021, as well as to reformulate it to elaborate new proposals for relevant H2020 calls involving AI-driven applications, especially LEIT-ICT, LEIT-NMBP or FET calls. Seeking to maximize the probability of success of such projects, the present proposal requests funding for (i) the composition of a compact and interdisciplinary consortium and the organization of a first in-person meeting, (ii) the writing of an executive summary of the project and subsequent discussions with a project officer, (iii) the definition of a responsibility assignment matrix and schedule to write the complete proposal, and (iv) the actual writing of the proposal. In all cases, preparations are based on the proposal grAIph submitted to the current ITN call (MSCA-ITN-2019) on January 25th, 2019 where Universitat Politècnica de Catalunya is the coordinator and Spanish organizations account for 38% of the total requested budget that is close to 4 million Euros. The project associated to the present proposal, grAIph, addresses one of the main outstanding challenges of Machine Learning (ML). ML has taken the world by storm and has become a fundamental pillar of engineering due to its capacity to solve extremely complex problems. However, and in spite of its all-pervasive applicability and potential, it is well-known that NOT all neural network architectures fit to all problems. Techniques such as Convolutional Neural Networks or Recurrent Neural Networks are tailored to model knowledge from the locality (e.g, images) and temporal sequentiality (e.g., language) of data respectively. As a result, existing ML techniques do not work well with relational or structured data, that is: graphs. However, graphs are the basic representation of information for many scientific and industrial fields such as neuroscience, systems biology or communication networks. In general, many fields are fundamentally based on complex networks and systems and, in order to take advantage of AI techniques there is a crucial need of ML techniques able to learn and model graphs. The acronym grAIph stands for Graph Artificial Intelligence for Biological and Communication Networks as we seek to enable a new breed of AI-powered applications in these fields through the use of graph AI. To this end, grAIph addresses the main issues related to graph AI nowadays: lack of experts, the disconnection between theory and practice, and the lack of training materials to engage the education and research community in this area. For this, grAIph implements a research and training program that covers theory, both in AI and complex systems as well as practice, with strong emphasis on neuroscience, systems biology and communication networks applications.
Plan Estatal de Investigación Científica y Técnica y de Innovación 2017-2020
Programa Estatal de I+D+i Orientada a los Retos de la Sociedad
Guirado, R.; Jain, A.; S. Abadal; Alarcon, E. IEEE International Symposium on Circuits and Systems p. 1-5 DOI: 10.1109/ISCAS51556.2021.9401612 Presentation's date: 2021-05 Presentation of work at congresses