The Big Data Era has posed new challenges to the scientific community, from the construction of computing infrastructure able to handle huge amounts of data, to the design of mathematical tools to represent and process big and heterogeneous data. A clever and energy-aware blend of HPC resources and scientific computing advances is key to designing successful and accurate responses in (almost) real time in a wide range of application settings. The work of computational mathematicians is thus crucial to properly deal with such urgent challenges. Recently explored strategies aim at saving computer energy by promoting the use of low cost operations on given computational architectures.

Randomized algorithms and mixed precision arithmetic are emerging as some of the most promising tools to boost numerical computations. The first approach (also called sketching) performs a dimensionality reduction while preserving the main features of the original problem with probabilistic confidence. The second approach aims at identifying the operations that can be carried out less accurately in a given algorithm and performing these operations using lower precision arithmetic, saving computations and computer power. Though of different nature, sketching and mixed precision approaches pursue the same goal, namely attaining a remarkable improvement in the efficiency of potentially any scientific computing algorithm.

We aim to gather practitioners and younger researchers to highlight the state of the art in both areas. The informal workshop environment will allow us to promote fruitful scientific discussions and initiate new collaborations among the participants from the two communities, on, e.g.,. combining sketching and mixed precision arithmetic schemes.

The two-day workshop will envision plenary and contributed talks followed by significant time windows for brainstorming and discussion.

Scientific and Organizing Committee

Davide Palitta
Alma Mater Studiorum Università di Bologna