Neural Simulation Tool Research Articles

Efficient parameter calibration and real-time simulation of large-scale spiking neural networks with GeNN and NEST.

Spiking neural networks (SNNs) represent the state-of-the-art approach to the biologically realistic modeling of nervous system function. The systematic calibration for multiple free model parameters is necessary to achieve robust network function and demands high computing power and large memory resources. Special requirements arise from closed-loop model simulation in virtual environments and from real-time simulation in robotic application. Here, we compare two complementary approaches to efficient large-scale and real-time SNN simulation. The widely used NEural Simulation Tool (NEST) parallelizes simulation across multiple CPU cores. The GPU-enhanced Neural Network (GeNN) simulator uses the highly parallel GPU-based architecture to gain simulation speed. We quantify fixed and variable simulation costs on single machines with different hardware configurations. As a benchmark model, we use a spiking cortical attractor network with a topology of densely connected excitatory and inhibitory neuron clusters with homogeneous or distributed synaptic time constants and in comparison to the random balanced network. We show that simulation time scales linearly with the simulated biological model time and, for large networks, approximately linearly with the model size as dominated by the number of synaptic connections. Additional fixed costs with GeNN are almost independent of model size, while fixed costs with NEST increase linearly with model size. We demonstrate how GeNN can be used for simulating networks with up to 3.5 · 106 neurons (> 3 · 1012synapses) on a high-end GPU, and up to 250, 000 neurons (25 · 109 synapses) on a low-cost GPU. Real-time simulation was achieved for networks with 100, 000 neurons. Network calibration and parameter grid search can be efficiently achieved using batch processing. We discuss the advantages and disadvantages of both approaches for different use cases.

Reproducing a decision-making network in a virtual visual discrimination task.

We reproduced a decision-making network model using the neural simulator software neural simulation tool (NEST), and we embedded the spiking neural network in a virtual robotic agent performing a simulated behavioral task. The present work builds upon the concept of replicability in neuroscience, preserving most of the computational properties in the initial model although employing a different software tool. The proposed implementation successfully obtains equivalent results from the original study, reproducing the salient features of the neural processes underlying a binary decision. Furthermore, the resulting network is able to control a robot performing an in silico visual discrimination task, the implementation of which is openly available on the EBRAINS infrastructure through the neuro robotics platform (NRP).

A Neuronal Network Model of the Primate Visual System: Color Mechanisms in the Retina, LGN and V1.

Color plays a key role in human vision but the neural machinery that underlies the transformation from stimulus to perception is not well understood. Here, we implemented a two-dimensional network model of the first stages in the primate parvocellular pathway (retina, lateral geniculate nucleus and layer 4C in V1) consisting of conductance-based point neurons. Model parameters were tuned based on physiological and anatomical data from the primate foveal and parafoveal vision, the most relevant visual field areas for color vision. We exhaustively benchmarked the model against well-established chromatic and achromatic visual stimuli, showing spatial and temporal responses of the model to disk- and ring-shaped light flashes, spatially uniform squares and sine-wave gratings of varying spatial frequency. The spatiotemporal patterns of parvocellular cells and cortical cells are consistent with their classification into chromatically single-opponent and double-opponent groups, and nonopponent cells selective for luminance stimuli. The model was implemented in the widely used neural simulation tool NEST and released as open source software. The aim of our modeling is to provide a biologically realistic framework within which a broad range of neuronal interactions can be examined at several different levels, with a focus on understanding how color information is processed.

Simulation of serotonin mechanisms in NEUCOGAR cognitive architecture

This work aims at demonstrating that the neuromodulatory mechanisms that control the emotional states of mammals (specifically rat’s brains) can be represented and re-implemented in a computational model processed by a machine. In particular we specifically focus on two neuro-transmitters, serotonin and dopamine, starting from their fundamental role in basic cognitive processes. In our specific implementation, we represent the simulation of the ‘disgust-like’ state based on the three dimensional neuromodulatory model of affects or emotions, according to the ‘cube of emotions’. These functional mechanisms can be transferred into an artificial cognitive system: inhibition, for example, can elicit a blocking behaviour that, depending on its intensity and duration, can push the system to a general emotional state. We have simulated 1000 milliseconds of the serotonin and dopamine systems using NEST Neural Simulation Tool with the rat brain as the model to artificially reproduce this mechanism on a computational system.

Emotional simulations and depression diagnostics

Abstract In this work we propose the following hypothesis: the neuromodulatory mechanisms that control the emotional states of mammals can be translated and re-implemented in a computer by controlling the computational performance of a hosted computational system. In our specific implementation, we represent the simulation of the ‘fear-like’ state based on the three dimensional neuromodulatory model of affects, in this paper ‘affects’ refer to the basic emotional inborn states, inherited from works of Hugo Lovheim. Whilst dopamine controls attention, serotonin is the key for inhibition, and fear is a elicitator for inhibitory and protective processes. This inhibition can promote [in a cognitive system] to blocking behaviour which can be labelled as ’depression’. Therefore, our interest is how to reimplement biomimetically both action-regulators without the computational system to resulting in a ‘failed’ scenario. We have simulated 1000 ms of the dopamine system using NEST Neural Simulation Tool with the rat brain as the model. The results of the simulation experiments are reported with an evaluation to demonstrate the correctness of our hypothesis.

The effect of longer range connections on neuronal network dynamics

The effect of longer range connections on neuronal network dynamics

Interactive visualization of brain-scale spiking activity

In recent years, the simulation of spiking neural networks has advanced in terms of both simulation technology [1,2] and knowledge about neuroanatomy [3,4]. Due to these advances, it is now possible to run simulations at the brain scale [5,6], which produce an unprecedented amount of data to be analyzed and understood by researchers. To aid computational neuroscientists with the development of models and especially with the visual inspection and selection of data for analysis, we developed VisNEST [7], a tool for the combined visualization of simulated spike data and anatomy. This provides a rapid overview of the relationship between structure and activity. VisNEST currently uses spike data from the neural simulation tool NEST [1] and geometry from the Scalable Brain Atlas [4], but is not limited to these tools. In our contribution we will present VisNEST using a Picasso 3D system, which allows users to interactively investigate and explore the simulated data from a large-scale model of 32 vision-related areas of the macaque [6]. The system is equipped with infrared tracking and uses passive glasses to render the image for the user standing in front of the screen. Figure 1 Main view of the simulated activity data. The mean spiking activity of the different areas is shown by color. The optional dot plot shows the spikes from the currently selected area.

A Python test suite for statistical properties of probabilistic networks with and without spatial structure

A Python test suite for statistical properties of probabilistic networks with and without spatial structure

Evaluating the Connection-Set Algebra for the neural simulator NEST

Event Abstract Back to Event Evaluating the Connection-Set Algebra for the neural simulator NEST Jochen Martin Eppler1*, Håkon Enger2, Thomas Heiberg2, Birgit Kriener2, Hans Ekkehard Plesser2, Markus Diesmann1 and Mikael Djurfeldt3 1 Research Center Juelich, Institute of Neuroscience and Medicine (INM-6), Germany 2 Norwegian University of Life Sciences, Mathematical Sciences and Technology, Norway 3 KTH, PDC Center for High Performance Computing, Sweden We present our ongoing work to evaluate the usability and efficiency of an implementation of the Connection Set Algebra (http://software.incf.org/software/csa) [1] for the neural simulation tool NEST (http://www.nest-initiative.org) [2]. Present day simulation languages work well for simple network structures like random or feed-forward networks. However, they either lack the required expressiveness or the required computational efficiency to specify the connectivity of complex hierarchical network models now under investigation. This increases the risk of errors in the model description or prevents the investigation of relevant model sizes. Therefore, a description language in the problem domain of the neuroscientist is required which guides the researcher to correct model specifications and avoids technical clutter. The Connection Set Algebra (CSA) is a generic framework, which provides elementary connection sets and operators for combining them. It allows one to define network connectivity in a compact and accessible format. A prototype implementation of the CSA exists in form of a Python module (libcsa). NEST is a simulator for large heterogeneous networks of point neurons or neurons with a small number of compartments. The concrete goals of this project are: Reduce the complexity of NEST's native connection routines through the use of CSA. Base the topology module of NEST on CSA in order to obtain better scaling, especially for very large networks on HPC facilities. Create a C++ implementation of CSA (libcsa) that can be used by neural simulators to describe network connectivity. The result of this study is a test implementation and a quantitative assessment of the proposed technology. Acknowledgments Partially funded by "The Next-Generation Integrated Simulation of Living Matter" project, part of the Development and Use of the Next-Generation Supercomputer Project of the Ministry of Education, Culture, Sports, Science and Technology (MEXT) of Japan, the Helmholtz Alliance on Systems Biology, the Research Council of Norway and the International Neuroinformatics Coordinating Facility.

PyNEST: A convenient interface to the NEST simulator

The neural simulation tool NEST (http://www.nest-initiative.org) is a simulator for heterogeneous networks of point neurons or neurons with a small number of compartments. It aims at simulations of large neural systems with more than 104 neurons and 107 to 109 synapses. NEST is implemented in C++ and can be used on a large range of architectures from single-core laptops over multi-core desktop computers to super-computers with thousands of processor cores. Python (http://www.python.org) is a modern programming language that has recently received considerable attention in Computational Neuroscience. Python is easy to learn and has many extension modules for scientific computing (e.g. http://www.scipy.org). In this contribution we describe PyNEST, the new user interface to NEST. PyNEST combines NEST's efficient simulation kernel with the simplicity and flexibility of Python. Compared to NEST's native simulation language SLI, PyNEST makes it easier to set up simulations, generate stimuli, and analyze simulation results. We describe how PyNEST connects NEST and Python and how it is implemented. With a number of examples, we illustrate how it is used.

PyNEST: a convenient interface to the NEST simulator

Event Abstract Back to Event PyNEST: a convenient interface to the NEST simulator Jochen M. Eppler1*, Marc-Oliver Gewaltig1, Moritz Helias2 and Eilif Muller3 1 Honda Research Institute Europe GmbH, Germany 2 BCCN Freiburg, Germany 3 Swiss Federal Institute of Technology, Switzerland Previous versions of NEST were difficult to use because simulations had to be formulated in a custom programming language (SLI), which is based on PostScript. This lead to a high barrier for new users. The upcoming version 2.0 of NEST lowers this barrier significantly by providing a convenient and easy-to-learn interface to NEST as an extension module for the Python programming language. This extension is called PyNEST. The neural simulation tool NEST (http://www.nest-initiative.org) is a simulator for heterogeneous networks of point neurons or neurons with a small number of compartments. It aims at simulations of large neural systems, i.e. networks with more than 10^4 neurons and up to 10^9 synapses. NEST is implemented as an efficient library in C++ that can be used on a large range of architectures ranging from desktop computers to large clusters with thousands of processor cores. Python (http://www.python.org) is a modern programming language that has recently gained considerable attention in Computational Neuroscience. Python is easy to learn and has many extension modules for scientific computing (http://www.scipy.org). PyNEST is an extension module for Python that combines the power of the NEST simulation kernel with the flexibility of Python. PyNEST allows faster simulation setup, easier construction of stimuli, and convenient analysis of the results in the same script. In this contribution we describe the main features of PyNEST from a user perspective. In a companion contribution we give a live demonstration of PyNEST. Conference: Neuroinformatics 2008, Stockholm, Sweden, 7 Sep - 9 Sep, 2008. Presentation Type: Poster Presentation Topic: Large Scale Modeling Citation: Eppler JM, Gewaltig M, Helias M and Muller E (2008). PyNEST: a convenient interface to the NEST simulator. Front. Neuroinform. Conference Abstract: Neuroinformatics 2008. doi: 10.3389/conf.neuro.11.2008.01.083 Copyright: The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers. They are made available through the Frontiers publishing platform as a service to conference organizers and presenters. The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated. Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed. For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions. Received: 25 Jul 2008; Published Online: 25 Jul 2008. * Correspondence: Jochen M Eppler, Honda Research Institute Europe GmbH, Offenbach, Germany, j.eppler@fz-juelich.de Login Required This action requires you to be registered with Frontiers and logged in. To register or login click here. Abstract Info Abstract The Authors in Frontiers Jochen M Eppler Marc-Oliver Gewaltig Moritz Helias Eilif Muller Google Jochen M Eppler Marc-Oliver Gewaltig Moritz Helias Eilif Muller Google Scholar Jochen M Eppler Marc-Oliver Gewaltig Moritz Helias Eilif Muller PubMed Jochen M Eppler Marc-Oliver Gewaltig Moritz Helias Eilif Muller Related Article in Frontiers Google Scholar PubMed Abstract Close Back to top Javascript is disabled. Please enable Javascript in your browser settings in order to see all the content on this page.

NEST (NEural Simulation Tool)

NEST (NEural Simulation Tool)

Exploring self‐organized emergence in an agent‐based synthetic warfare lab

Artificial‐life techniques – specifically, agent‐based models and evolutionary learning algorithms – provide a potentially powerful new approach to understanding some of the fundamental processes of war. This paper introduces a simple artificial‐like “toy model” of combat called Enhanced ISAAC Neural Simulation Tool (EINSTein). EINSTein is designed to illustrate how certain aspects of land combat can be viewed as self‐organized, emergent phenomena resulting from the dynamical web of interactions among notional combatants. EINSTein's bottom‐up, synthesist approach to the modeling of combat stands in stark contrast to the more traditional top‐down, or reductionist approach taken by conventional military models, and represents a step toward developing a complex systems theoretic toolbox for identifying, exploring, and possibly exploiting self‐organized emergent collective patterns of behavior on the real battlefield. A description of the model is provided, along with examples of emergent agent patterns and behaviors.

Dynamic simulation of a steam generator by neural networks

Numerical simulation by computers of the dynamic evolution of complex systems and components is a fundamental phase of any modern engineering design activity. This is of particular importance for risk-based design projects which require that the system behavior be analyzed under several and often extreme conditions. The traditional methods of simulation typically entail long, iterative, processes which lead to large simulation times, often exceeding the transients real time. Artificial neural networks (ANNs) may be exploited in this context, their advantages residing mainly in the speed of computation, in the capability of generalizing from few examples, in the robustness to noisy and partially incomplete data and in the capability of performing empirical input–output mapping without complete knowledge of the underlying physics. In this paper we present a novel approach to dynamic simulation by ANNs based on a superposition scheme in which a set of networks are individually trained, each one to respond to a different input forcing function. The dynamic simulation of a steam generator is considered as an example to show the potentialities of this tool and to point out the difficulties and crucial issues which typically arise when attempting to establish an efficient neural network simulator. The structure of the networks system is such to feedback, at each time step, a portion of the past evolution of the transient and this allows a good reproduction of also non-linear dynamic behaviors. A nice characteristic of the approach is that the modularization of the training reduces substantially its burden and gives this neural simulation tool a nice feature of transportability.

Using a general-purpose neural network simulation tool—NEUCOMP—for character recognition problems

NEUCOMP is a neural network compiler that compiles a program written as a list of mathematical specifications of neural network (NN) models and then translates it into a chosen target program. A simulation program written in the NEUCOMP language based on selected NN models for solving character recognition problems is presented. The models are the Multilayer, Counterpropagation, Kohonen self-organizing networks, and Adaptive Resonance Theory (ART), which are chosen for their differences in architecture and learning algorithms. The program runs on the Sequent Balance machine at PARC.