Elsevier

Neural Networks

Volume 24, Issue 9, November 2011, Pages 918-926
Neural Networks

2011 Special Issue
PLATO: Data-oriented approach to collaborative large-scale brain system modeling

https://doi.org/10.1016/j.neunet.2011.06.011Get rights and content

Abstract

The brain is a complex information processing system, which can be divided into sub-systems, such as the sensory organs, functional areas in the cortex, and motor control systems. In this sense, most of the mathematical models developed in the field of neuroscience have mainly targeted a specific sub-system. In order to understand the details of the brain as a whole, such sub-system models need to be integrated toward the development of a neurophysiologically plausible large-scale system model. In the present work, we propose a model integration library where models can be connected by means of a common data format. Here, the common data format should be portable so that models written in any programming language, computer architecture, and operating system can be connected. Moreover, the library should be simple so that models can be adapted to use the common data format without requiring any detailed knowledge on its use. Using this library, we have successfully connected existing models reproducing certain features of the visual system, toward the development of a large-scale visual system model. This library will enable users to reuse and integrate existing and newly developed models toward the development and simulation of a large-scale brain system model. The resulting model can also be executed on high performance computers using Message Passing Interface (MPI).

Highlights

► PLATO is a data-oriented approach for large-scale system modeling. ► We develop a model integration library based on the PLATO’s concept. ► A common data format applies for exchanging data between models. ► Models can plug into a large-scale model, just like playing with LEGO blocks. ► PLATO ensures high performance computing utilizing message passing interface (MPI).

Introduction

The brain can be considered as a large-scale information processing  system, which flexibly performs functions such as recognition, perception, learning, memory, and motor control. It receives information from the external world through sensory systems, processes them based on learned memory, and generates the output motor commands through motor control systems. The elucidation of the complicated information processing underlying those functions is extremely important. In order to grasp the mechanisms underlying such functions, numerous neurophysiological studies have been conducted, and models describing individual features have been developed. To uncover the complicated information processing in the whole brain system, detailed models of each sub-system should be constructed and integrated. To this end, several projects such as those reviewed by de Garis, Shuo, Goertzel, and Ruiting (2010) and Goertzel, Ruiting, Arel, de Garis, and Shuo (2010) were attempted to develop large-scale models that target specific brain functions and system models that integrate sub-system models developed by different research groups. However, in practice, models are described in different levels (i.e., functional, computational, realistic, etc.), programming languages, and data structures. Even when dealing with the same function or object, the model’s input, output, and parameters may have different formats. For this reason, even if two models are described using the same simulator or language, its integration becomes very complicated, and much effort may be required for adapting the codes to each other.

As an approach for large-scale mathematical modeling and simulation studies in the field of neuroscience, one could use systematic modeling languages and simulators such as NEURON (Hines & Carnevale, 1997), GENESIS (Bower & Beeman, 1997), and NEST (Gewaltig & Diesmann, 2007), where neural simulations are executed by simply describing the structure of the network and model parameters. On the other hand, XML1 document type modeling languages such as CellML (Hedley, Nelson, Bellivant, & Nielsen, 2001), SBML (Hucka et al., 2003), NeuroML (Gleeson et al., 2010), InsilicoML (Asai et al., 2008), and NineML (Gorchetchnikov and INCF Multiscale Modeling Taskforce, 2010, Raikov and INCF Multiscale Modeling Taskforce, 2010) are specialized in describing the formula, parameter, and structure of a network, and are separated from simulation techniques and algorithms. In this case, simulation programs interpret the XML document and either simulate them or translate them to run on the simulators described above. Models are ported for simulators using the same XML document; thus, integration can be carried out within the same simulator.

Based on the aforementioned complications involved in integrating large-scale models, we have recently developed a novel modeling framework called PLATO (platform for a collaborative brain system modeling) (Usui, Inagaki et al., 2009, Usui, 2010). The objective of PLATO is to enable the reuse and integration of existing models stored in neuroinformatics databases (Hines et al., 2004, Migliore et al., 2003, Usui et al., 2008, Usui and Okumura, 2008) in order to construct and simulate a large-scale model. Newly developed models can also be integrated into an existing large-scale model. As an example, we are developing a large-scale model of the whole visual system to understand the visual processing underlying perception, illusion, learning, and memory.

In this paper, we describe a data-oriented model integration method in PLATO. A common data format (see below) is utilized for the model and/or sub-model (sub-system model) interface for exchanging data between models. An agent process manages the entire simulation by controlling the timing of model execution and data exchange. In the simulation, our approach successfully connected different types of models described by C++ and Python with small changes in program code.

Section snippets

Data-oriented model integration

The main idea underlying the data-oriented model integration method is to commonize the models’ input and output (I/O) data so that different models can be connected to each other. This provides a framework for model connection toward the development of a large-scale model from existing and newly developed sub-models. It also allows sub-models to be replaced so that the large-scale model can be improved as more detailed sub-models are developed. In other words, sub-models can be plugged into

Data convention

To create a pluggable model without deep knowledge on the models to be connected, the configuration of the model input and output data should be designed ahead of the development of the model. The description of the contents of a data file in netCDF uses an XML document based on the NcML Schema4 and is called “Data Convention”. This document describes three types of information: “Dimension”, “Variable”, and

An example of visual system

In order to demonstrate the usefulness of the proposed data-oriented model integration library, we constructed a visual system model composed of an eye movement model (Inagaki, Hirata, & Usui, 2011), an eye optics model (unpublished), a pupil model, and a retinal network model named “VirtualRetina” (Wohrer & Kornprobst, 2009). Fig. 5(A) illustrates a schematic diagram of the visual system model constructed using PLATONIC. The eye optics model was constructed based on Artal’s model (Artal, 1990)

Contributions to large-scale modeling

In large-scale simulations, specifications and limitations are often incorporated to optimize efficiency. The development of a large-scale model may be hindered by such specifications and limitations. PLATONIC tries to develop a large-scale system model by loosely coupling existing sub-system models. That is, the principal objective of PLATONIC is to connect the models, rather than considering the large-scale system performance.

When users develop models adapted to PLATONIC, the models’ I/O data

Conclusions

This paper presented a data-oriented model integration library as part of the PLATO framework that aims to integrate and develop a large-scale mathematical model of the brain. The proposed data-oriented model integration library, called PLATONIC, employs TCP and MPI for communication in the model simulation. It provides a framework for environment-independent, flexible, and scalable model integration by isolating the functions of data I/O, management of simulation time, and calculations. This

Acknowledgments

We thank Drs. Shunji Satoh, Yoshimi Kamiyama, Yutaka Hirata, Akito Ishihara, and Hayaru Shouno for valuable discussion for development and Yoshihiro Okumura for technical support. This work is 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 of Japan.

References (30)

  • A. Gorchetchnikov et al.

    NineML—a description language for spiking neuron network modeling: the user layer

    BMC Neuroscience

    (2010)
  • W. Gropp et al.

    Using MPI: portable parallel programming with the message-passing interface

    (1994)
  • W.J. Hedley et al.

    A short introduction to CellML

    Philosophical Transactions of Royal Society A

    (1783)
  • M.L. Hines et al.

    The NEURON simulation environment

    Neural Computation

    (1997)
  • M.L. Hines et al.

    ModelDB: a database to support computational neuroscience

    Journal of Computational Neuroscience

    (2004)
  • Cited by (7)

    • Reprint of: Simulation Platform: A cloud-based online simulation environment

      2011, Neural Networks
      Citation Excerpt :

      The latter is composed of eight databases and is organized by Neuroinformatics Japan Center, which is the Japan Node (J-Node) of the International Neuroinformatics Coordinating Facility. J-Node Platforms serve as online databases, and a companion project PLATO (Kannon, Inagaki, Kamiji, Makimura, & Usui, 2011; Usui, 2010) aims to provide the scheme of model integration. For the integration, we have to validate that all component models of an integrated model work properly beforehand.

    • Simulation Platform: A cloud-based online simulation environment

      2011, Neural Networks
      Citation Excerpt :

      The latter is composed of eight databases and is organized by Neuroinformatics Japan Center, which is the Japan Node (J-Node) of the International Neuroinformatics Coordinating Facility. J-Node Platforms serve as online databases, and a companion project PLATO (Kannon, Inagaki, Kamiji, Makimura, & Usui, 2011; Usui, 2010) aims to provide the scheme of model integration. For the integration, we have to validate that all component models of an integrated model work properly beforehand.

    • Collaborative modelling: The future of computational neuroscience?

      2012, Network: Computation in Neural Systems
    View all citing articles on Scopus
    View full text