Search

Running scientific workflows on a supercomputer can be a daunting task for a scientific domain specialist. Workflow management solutions (WMS) are a standard method for reducing the complexity of application deployment on high performance computing (HPC) infrastructure. We introduce the design for a middleware system that extends and combines the functionality from existing solutions in order to create a high-level, staged user-centric operation/deployment model. This design addresses the requirements of several use cases in the life sciences, with a focus on neuroscience. In this manuscript we focus on two use cases: 1) three coupled neuronal simulators (for three different space/time scales) with in-transit visualization and 2) a closed-loop workflow optimized by machine learning, coupling a robot with a neural network simulation. We provide a detailed overview of the application-integrated monitoring in relationship with the HPC job. We present here a novel usage model for large scale interactive multi-application workflows running on HPC systems which aims at reducing the complexity of deployment and execution, thus enabling new science., Comment: 7 pages, 3 figures, The 2019 International Conference on High Performance Computing & Simulation

Individualized anatomical information has been used as prior knowledge in Bayesian inference paradigms of whole-brain network models. However, the actual sensitivity to such personalized information in priors is still unknown. In this study, we introduce the use of fully Bayesian information criteria and leave-one-out cross-validation technique on the subject-specific information to assess different epileptogenicity hypotheses regarding the location of pathological brain areas based on a priori knowledge from dynamical system properties. The Bayesian Virtual Epileptic Patient (BVEP) model, which relies on the fusion of structural data of individuals, a generative model of epileptiform discharges, and a self-tuning Monte Carlo sampling algorithm, is used to infer the spatial map of epileptogenicity across different brain areas. Our results indicate that measuring the out-of-sample prediction accuracy of the BVEP model with informative priors enables reliable and efficient evaluation of potential hypotheses regarding the degree of epileptogenicity across different brain regions. In contrast, while using uninformative priors, the information criteria are unable to provide strong evidence about the epileptogenicity of brain areas. We also show that the fully Bayesian criteria correctly assess different hypotheses about both structural and functional components of whole-brain models that differ across individuals. The fully Bayesian information-theory based approach used in this study suggests a patient-specific strategy for epileptogenicity hypothesis testing in generative brain network models of epilepsy to improve surgical outcomes., Author summary Reliable prediction of the Epileptogenic Zone (EZ) is a challenging task due to nontrivial brain network effects, non-linearity involved in spatiotemporal brain organization, and uncertainty in prior information. Based on the whole-brain modeling approach, the anatomical information of patients can be merged with a generative model of epileptiform discharges to build a personalized large-scale brain model of epilepsy spread. Here, we apply information criteria and cross-validation technique to a whole-brain model of epilepsy spread to infer and validate the spatial map of epileptogenicity across different brain areas. By definition, classical information criteria are independent of prior information, in which the penalty term (number of parameters and observed data) is the same across different EZ candidates, making them infeasible to determine the best among a set of epileptogenicity hypotheses. In contrast, the fully Bayesian information criteria and cross-validation enable us to integrate our prior information to improve out-of-sample prediction accuracy for EZ identification. Using the dynamical system properties of a whole-brain model of epilepsy spread, and dependent on the level of prior information, the proposed approach provides accurate and reliable estimation about the degree of epileptogenicity across different brain areas. Our fully Bayesian approach relying on automatic inference suggests a patient-specific strategy for EZ prediction and hypothesis testing before therapeutic interventions.

Modeling large-scale spiking neural networks showing realistic biological behavior in their dynamics is a complex and tedious task. Since these networks consist of millions of interconnected neurons, their simulation produces an immense amount of data. In recent years it has become possible to simulate even larger networks. However, solutions to assist researchers in understanding the simulation's complex emergent behavior by means of visualization are still lacking. While developing tools to partially fill this gap, we encountered the challenge to integrate these tools easily into the neuroscientists' daily workflow. To understand what makes this so challenging, we looked into the workflows of our collaborators and analyzed how they use the visualizations to solve their daily problems. We identified two major issues: first, the analysis process can rapidly change focus which requires to switch the visualization tool that assists in the current problem domain. Second, because of the heterogeneous data that results from simulations, researchers want to relate data to investigate these effectively. Since a monolithic application model, processing and visualizing all data modalities and reflecting all combinations of possible workflows in a holistic way, is most likely impossible to develop and to maintain, a software architecture that offers specialized visualization tools that run simultaneously and can be linked together to reflect the current workflow, is a more feasible approach. To this end, we have developed a software architecture that allows neuroscientists to integrate visualization tools more closely into the modeling tasks. In addition, it forms the basis for semantic linking of different visualizations to reflect the current workflow. In this paper, we present this architecture and substantiate the usefulness of our approach by common use cases we encountered in our collaborative work.

We introduce Arbor, a performance portable library for simulation of large networks of multi-compartment neurons on HPC systems. Arbor is open source software, developed under the auspices of the HBP. The performance portability is by virtue of back-end specific optimizations for x86 multicore, Intel KNL, and NVIDIA GPUs. When coupled with low memory overheads, these optimizations make Arbor an order of magnitude faster than the most widely-used comparable simulation software. The single-node performance can be scaled out to run very large models at extreme scale with efficient weak scaling. HPC, GPU, neuroscience, neuron, software, Comment: PDP 2019 27th Euromicro International Conference on Parallel, Distributed and Network-based Processing