CalsimLab objectives are to build a coherent theoretical background, develop adapted numerical methods and implement efficient algorithms, for the following four Grand Challenges:
Linear Scaling in Computational Chemistry will address the need to design and implement a reliable and efficient procedure that takes into account electronic correlation and would scale almost linearly with the system size.
Molecule Energy Approximation in Computational Chemistry will overcome the simpler approximations of the Schrödinger equations that have been introduced by chemists to estimate the energy of molecules in various states (from ground states to excited states).
Sequential Algorithms in Computational Biology will address the major hurdle of improving sequential algorithms that are significantly difficult to parallelise and adapt.
Algorithms for Genomics will implement multivariate models in integrative genomics for ultra-high dimensional data and will couple them with inferential algorithms that search over vast model spaces.
The underlying CalsimLab vision is to transform the existing environment for scientific computing at UPMC into a new structure that will boost dissemination and increase expertise. The aim is to achieve prime research objectives more efficiently and in all disciplines where modelling and simulation are mandatory. CalsimLab will create a new research framework to take up existing and emerging challenges and demonstrate its efficiency by applying it first to computational chemistry and computational biology.
Modelling and simulation to understand mechanisms and test hypothesis in research Modelling and simulation is becoming a new discipline with manifold goals. The aims are to increase the understanding of systems, and of their associated models, to improve the system/model performances and, to test and check new concepts and new hypotheses.
The consistent combination of modelling, theoretical analysis, simulation and experimental validation is the key principle behind scientific computing. This requires the establishment of an on-going close scientific relationship between all disciplines involved which will encourage, strong interdisciplinary cooperation.
Bridging the knowledge gap between disciplines The advent of computers has led mathematicians working on models to collaborate with computer scientists working on algorithms. This has yielded solutions to important practical problems for physicists. Historically, mathematicians and computer scientists have worked more closely with engineers and physicists than with scientists from other disciplines, for example chemists or biologists.
Nowadays, all disciplines concerned with scientific computing are already well aware that mutual interplay is bringing new challenges that are no longer contained within their own fields of expertise and that a large part of their research activity lies at the interface between different scientific fields.
CalsimLab proposal The ultimate goal of CalsimLab is to contribute to scientific breakthroughs at UPMC units whenever scientific computing and simulation is involved. This long-term goal implies the development of new and better theoretical methods and computational codes in many fields. Indeed, as pointed out by leading experts, “research harvesting is possible by using other programs, but it is an abortive flower”. True breakthroughs can only be achieved if the researchers are not circumscribed by the current software limitations.
The availability of long-term support will enable the building of new packages from scratch to fit both competitive scientific issues and upcoming architecture. CalsimLab partners are taking up the challenge of developing new models, new numerical schemes and building new state-of-the-art packages.
CalsimLab focuses on two research areas to demonstrate the feasibility of the approach:
The proposal is to engage the relevant scientific communities at UPMC in a roadmap to identify and to address the opportunities and the challenges ahead in these two disciplines that are highly concerned with scientific computing and with computing at an extreme scale. Moreover, the two disciplines share a common interest in quantum modelling, although with different numerical approaches and HPC usage, and therefore will benefit from the synergies created within the proposal.