Resumen
The use of machine learning approaches in science and technology research is ubiquitous nowadays. This is due to the
combination of three situations: the increasing amount of available data, the increasing computing capability, and the variety
of different models and paradigms able to learn from data. Among the last ones, we can distinguish two very different
models: those which consist on black box systems, like (deep) neural networks, and those that once trained allow to extract
information that explains how they operate and what have they learned from data [1-3]. While the former are very efficient
and in providing solutions to particular problems, they have two main drawbacks from our point of view: it is difficult to
guarantee their performance in different or varying contexts and they provide no (or limited) insight about how the data
should be processed and about how inputs and outputs are related. The latter models are indeed designed from a different
approach. They are able to learn from data and they provide insight about how to relate inputs and outputs and, if properly
interpreted, they can even provide knowledge about the phenomenon related to the data that can in turn be used to
understand it better and to develop models for it and really contribute to the development of related science and technology.
In this context, we propose this project with the objective of using interpretable machine learning methods in different
problems related to vision, imaging and color sciences with the objective of not only obtaining systems able to solve specific
problems in each of the fields but also to analyze what the systems have learned in order to interprete the knowledge that
has been extracted from data and to study whether this knowledge can be included in current scientific models in the state of
the art for the corresponding problems, an approach that has led to interesting results so far [4].
For the proper development of the project we count with a multidisciplinary team with knowledge in mathematical modelling,
vision, color and imaging sciences and interpretable artificial intelligence models. But also the team needs to have access to
both data resources as well as laboratories where more data can be obtained if needed, which we do as well. Most of the
team has been working together for a long time, where we have obtained relevant scientific developments that have resulted
in numerous publications in highly prestigious journals and conference
