M4 Digital Society
The widespread and massive access to the internet (and in particular the “internet of things”), the implantation of sensors in all areas of life (both personal and professional) as well as in industrial activity, the robotization of industry and the automation of routine tasks has led us to live in a digitized and digital society, in which the information generated and stored far exceeds the current capacity for analysis.
On the other hand, the growing success of artificial intelligence, due in large part to the capacity of specialized processors, necessitates the search for mathematical foundations to understand the inner workings of algorithms and thus provide a basis for improving or making them more precise and predictable.
Our contributions in this area will be organized into three PIs.
PI Intelligent systems: Given the enormous amount of data at our fingertips (“big data”) we have to move forward to find efficient ways to store it in the long term, analyze it and manage it. We also need to develop new ways of discriminating those that are important from those that are not and describe ways of acting on that data (optimization and control). In many cases, the knowledge obtained will have to be applied in real time, which requires the implementation of efficient algorithms. Finally, we need to reflect on how the use of this data affects and changes society, including pressing moral issues such as those related to privacy or human rights.
PI Digital Humanities: Museums and archives are making a considerable effort to digitize their material on different media, such as audio, video, text, and images. These materials are primary sources of information to answer new questions on a wide spectrum of disciplines such as history, language analysis, linguistic and literature studies, among others. The structures that generate the available data, as well as the dynamics followed by the processes involved, require the design of mathematical models that consider the multiple sources of variability and the implementation of dimension reduction methods to understand, for example, language variation and dialectology, or corpus and lexicographic analysis.
PI Computing, learning and security: The need to optimally determine a set of parameters is inherent in all machine learning methods. The underlying optimization problems vary according to the models and fields of application. Therefore, being able to define and / or identify to later implement the correct optimization algorithm is essential in the use of the increasingly complex machine learning models. The need for privacy and security has given rise to areas such as privacy-preserving data mining and encrypted computing, in which it is sought to be able to analyze a data set without compromising privacy and to be able to perform calculations on a data set while keeping it encrypted.