Advances in computational science, data management and analysis have been a key factor in the success of Brookhaven Lab's scientific programs at the Relativistic Heavy Ion Collider (RHIC), the National Synchrotron Light Source (NSLS), the Center for Functional Nanomaterials (CFN), and in biological, atmospheric, and energy systems science, as well as our collaborative participation in international research endeavors, such as the ATLAS experiment at Europe's Large Hadron Collider.
The Computational Science Initiative (CSI) brings together under one umbrella the expertise that has driven this success to foster cross-disciplinary collaborations to address the next generation of scientific challenges posed by facilities such as the new National Synchrotron Light Source II (NSLS II). A particular focus of CSI's work will be the research, development and deployment of novel methods and algorithms for the timely analysis and interpretation of high volume, high velocity, heterogeneous scientific data created by experimental, observational and computational facilities to accelerate and advance scientific discovery. CSI is hereby taking an integrated approach, providing capabilities from leading edge research to multi-disciplinary teams that deliver operational data analysis capabilities to the scientific user communities.
Computer Science and Mathematics—fundamental research into novel methods and algorithm in support of large-scale, multi-modal, and streaming data analysis. Novel solutions for long term data curation and active reuse. Approaches to enable energy efficient, extreme-scale numerical modeling specifically in computational materials science, chemistry, lattice quantum chromo dynamics and fusion.
The BNL Scientific Data and Computing Center, housing the latest systems in high-performance and data-intensive computing, data storage, and networking, offering everything from novel research platforms to highly reliable production services.
The Computational Science Laboratory, a collaborative space for the development of advanced algorithms and their characterization and optimization, also brings together computer scientists, mathematicians, and leading computational scientists to develop next-generation numerical simulation models
The Center for Data Driven Discovery (C3D), a multi-disciplinary center for the development, deployment, and operation of data-intensive discovery services for science, national security, and industry
The CSI philosophy is a multi-disciplinary and collaborative approach to scientific research and development, with research targeted at and informed by the key challenges observed in close interactions with our clients in science, national security agencies, and industry. Our success is measured in equal parts by the advancement we can bring to computer science and mathematics, as well as by the transformational impact we have on our clients’ mission space.
The CSI brings together under one umbrella the expertise that fosters cross-disciplinary collaboration and makes optimal use of existing technologies, while also leading the development of new tools and methods that will benefit science both within and beyond the Laboratory. Key partners include nearby universities such as Columbia, Cornell, New York University, Stony Brook, and Yale, as well as IBM Research.
Computational scientists at Brookhaven will also seek to establish partnerships with key players in academia and industry (e.g. Stony Brook University’s Institute for Advanced Computational Science, Rensselaer Polytechnic Institute, Oak Ridge National Laboratory, IBM, and Intel). One existing example of a successful partnership is the collaboration of Brookhaven Lab’s high-energy and nuclear physics research groups with IBM that led to the development of the BlueGene supercomputing architecture now used on the world’s most powerful commercially available supercomputers.More about strategic partnerships
Computational Science Initiative Event
"CSI Seminar: Recurrent Networks and Corresponding Applications"
Presented by Lin SUn, Stanford University
11 am, Seminar Room, Bldg. 725
Monday, March 27, 2017, 11:00 am
Hosted by: ''Kerstin Kleese van Dam''
Currently, the most successful learning models are based on the paradigm of successive learning of representations followed by a decision layer. This is most commonly actualized through feedforward multilayer neural networks, such as Convolution Neural Networks (CNNs), where each layer forms one of such successive representations. However, an alternative that can achieve the same goal is a recurrent networks or feedback networks, in which the representation is formed in an iterative manner according to a feedback received from previous iteration's outcome. In this talk, we will deeply investigate the feedback networks, particularly the convolutional LSTM which illustrates several fundamental advantages over feedforward: it enables making early predictions at the query time, its output conforms to a hierarchical structure in the label space (e.g. a taxonomy), and it provides a new basis for Curriculum Learning. We put forth a working feedback based learning architecture for image classification task on par or better than existing feedforward networks with the addition of the above advantages. We further investigate its effectiveness on human pose estimation and action recognition problems.