Keynote, invited and industry talks presented at CARLA 2019. Use each panel to read the abstract.
Keynote talks
Simplifying the development of complex workflows for distributed computing infrastructures — Rosa M. Badia
Distributed computing infrastructures are becoming more complex, with environments that involve sensors, edge devices, instruments and high end systems such as clouds and HPC clusters. A key aspect is how to describe and develop the applications executed on such platforms. Very often these applications are not standalone, but involve a set of sub applications or steps composing a workflow, with components of different nature that can combine computationally intensive and data analytics parts. COMPSs is a task based programming model that enables the development of workflows that can be executed in parallel on distributed computing platforms. The talk presents an overview of the challenges of workflow development in these environments and how they can be tackled with COMPSs.
Recognizing the flora of the world: dream or reality? — Hervé Goëau
Plant diversity plays a crucial role in the functioning and stability of ecosystems, yet our knowledge of plants is far from complete given nearly 400,000 known species and more than 2,000 discovered each year. Automated identification has made considerable progress, largely thanks to convolutional neural networks, as evidenced by the long term evaluation of automated plant identification organized within the LifeCLEF initiative. The best systems can now compete with human experts, and plant identification applications such as Pl@ntNet are used by a large audience, offering an opportunity to monitor biodiversity at a large scale. The talk discusses the possibilities and limitations of automated plant identification and how they can support biodiversity monitoring.
Promoting collaborations in HPC throughout Latin America: the case of SDumont and SINAPAD — Antônio Tadeu A. Gomes
The Santos Dumont petascale supercomputer (SDumont) is the largest HPC facility in Latin America dedicated to science. It is housed at the National Laboratory for Scientific Computing (LNCC) in Petropolis, Brazil, and is used by more than 900 users and 120 research projects across more than 20 research areas. The talk introduces the supercomputer and some of the research projects carried out on it, then discusses possibilities for collaboration between Latin American researchers on the use of this system.
Invited talks
Performance driven floating point tuning for GPU applications — Ignacio Laguna
The talk presents a static and dynamic analysis framework to perform mixed precision floating point tuning on scientific GPU applications. While precision tuning techniques are available, they are designed for serial programs and are accuracy driven, meaning they consider configurations that satisfy accuracy constraints but may degrade performance. The framework instead presents a performance driven approach for tuning floating point programs, obtaining performance improvements on several GPU applications.
New systems, new projects: the PSC HPC, HPAI and data journey — Paola Buitrago
Research is rapidly evolving to be even more data centric, with artificial intelligence driving the change and increasingly enabling qualitative breakthroughs. To bring these capabilities to the research community, the Pittsburgh Supercomputing Center designs and operates advanced computing systems that bring together HPC, AI and Big Data. The talk surveys recent PSC systems, its Open Compass research program and its life science data initiatives.
Computational protein 3D structure modeling from cryo electron microscopy density maps — Daisuke Kihara
The significant progress of cryo electron microscopy poses a pressing need for software for structural interpretation of EM maps, particularly for protein structure modeling at around 4 Angstrom resolution or worse. The talk introduces two tools from the speaker's laboratory, MAINMAST for de novo main chain tracing and Emap2sec, which uses a convolutional deep neural network for detecting protein structures.
Deep learning and mathematical morphology for biomedical applications — Etienne Decencière
In biomedical image segmentation, a priori information on the objects of interest is often available. This information can be difficult to use with deep learning methods, whereas mathematical morphology allows the information to be modeled in a natural manner. The talk proposes several ways of combining both approaches to solve practical applications in retinal and histological image analysis.
Complex networks of miRNA transcription factors mediating gene dosage compensation — Rodrigo Mora
Genomic instability leads to aneuploidy in most cancers. The talk explores the hypothesis that dosage compensation of key genes is a mechanism to deal with aneuploidy in cancer, examining genomic and transcriptomic data and identifying a cluster of genes with low variation in expression despite high variation in copy number.
Industry talks
Current challenges and trends in the design of HPC systems — Albert Trill
The talk presents current challenges and trends in HPC systems. Aspects such as in memory computing, Big Data and Artificial Intelligence, and cooling constraints greatly affect design, while the new paradigm of quantum computing introduces a radical effect.
Dell EMC HPC solutions portfolio: making innovation real — Joaquim Munoz Merino
An industry talk on the convergence of HPC and AI and the solutions portfolio that supports it, delivered by Dell Technologies.
The journey of memory driven computing — Fabio Alves
An industry talk from Hewlett Packard Enterprise on memory driven computing and its role in future HPC systems.
The convergence of HPC and AI: how Lenovo is addressing this trend — Jaime Puente
The supercomputing field is taking the next step in its evolutionary path, with researchers and enterprises increasingly applying HPC and AI for scientific insight. The talk presents how Lenovo approaches the convergence of HPC and AI, showcasing high impact research and enterprise projects.