Data analysis tools and model representation are a fascinating challenge in the age of the Internet of Data (IoD). Novel knowledge in IoD can be handled by adopting mathematical models, accurate and eﬃcient numerical methods, supported by promising learning approaches. Data representation using reduced space and the model approximation, using deep learning methodologies, could be effective solutions. This proposal aspires to exploit advanced nonlinear approximation techniques to cope with novel multidimensional data structures consisting of time series or high-dimensional manifold-valued data. Finally, it desires to bring a novel set of tools to the fascinating world Data Science research field.
Topics include, but are not limited to, following:
Multidimensional data approximation in Data Science contexts
Learning approaches in data representation
Numerical approximation by mean Kernel methods
Kernel methods and its applications in Data Science
Linear and non-Linear methodologies in Deep Learning
Signal processing through Data Science methodologies
Approximation in high-dimensional spaces
Predictive techniques for multidimensional time series in Data Science.