Dear Colleagues,

Welcome to the NeuroCAD project homepage!

introduction - history - license - support


Introduction

The NeuroCAD is free toolkit to design models and perform high speed computing of large biologically plausible neural networks(*).
Currently, NeuroCAD project consists of:

  1. NeuroCAD Simulation Engine
  2. Modules Repository
  3. Assemblers Repository
It is distributed under latest GNU General Public License. The basic goal of the project is to combine the universality of script machines like GENESIS and NEURON and the performance of highly optimized application. The main features of this project are listed as follow:
  1. Independence. The Simulator Engine is independent from any other part of NeuroCAD project. It also may be included in other projects as library;
  2. Maximal efficiency. The NeuroCAD Simulator Engine performs simulations as fast as possibly;
  3. Modularity. All model's primitives (ion channels, cells, synapses etc.) are implemented as separated modules which are used as building blocks to create any model of single neuron or neural network without any limitations. They are collected in shared Modules Repository;
  4. Hierarchy. All parts of NeuroCAD project are based on the NeuroCAD Simulator Engine and Modules Repository;
  5. Usability. All modules are written in classical C. A model of neural networks (Model Descriptor) is described as "assembling circuit" by general-purpose markup language(**);
  6. Synthesis. The runtime copy of the model is synthesized by Assembler from modules according to rules which are described in Model Descriptor every time before a simulation starts up;
  7. Extensibility. It is easy to create new module which can be used by NeuroCAD environment without any restrictions. It is also possible to create user-defined Assembler, to import specifically described model to NeuroCAD environment(***);
  8. Accessibility. Any model's parameter or variable can be easily controlled and displayed by NeuroCAD environment during simulation;
  9. Flexibility. NeuroCAD environment allows to implement different numerical methods and use them together in order to provide the most efficient way to compute of large biological plausible neural networks;
  10. Scalability. NeuroCAD environment provides efficient way to implement massive parallel computing to achieve maximal performance of simulations of large biologically plausible neural networks(****);
  11. Remote computing. NeuroCAD environment also provides a possibility to implement client/server paradigm that allows to compute models of neural networks remotely.

NeuroCAD is young developing project. In the very near future it will be posted at sourceforge.org and currently may be find at preliminary web page: http://nisms.krinc.ru/neurocad.org

(*) It also may be used as general purpose solver. The NeuroCAD environment provides a possibility to resolve a non-linear system of algebraic, differential, and integral equations together.)
(**) Authors chose XML as a language to describe model of neural networks.
(***) User-defined Assemblers allow to use different formats of Model Descriptor files, for example, it is possible to use models which are based on NEURON's or GENESIS's scripts as a native ones for NeuroCAD environment.
(****) Library for parallel computing (libpnce) currently under reconstructions and will be return to NeuroCAD project tree soon.




History

NeuroCAD porject is originally designed and maintained by Ruben Tikidji - Hamburyan and Sergey Markin (Lab. Neuroinformatic sensory and motor system (NISMS), of A.B. Kogan research institute for neurocybernetics (KRINC), Rostov State University.
With our previous experience in designing of applications for neuromodeling [Sergey Markin - SpCord (population network with H.-H. models of single cells); Ruben Tikidji - Hamburyan - NeuroCAD 2.01 old version for Windows(c) (large networks based on modified "Integrate - and - Fire" neurons)] we tried to find an universal and effective solution for modeling neural networks. An ultimat goal of NeuroCAD project is to provide an efficiency of 'solidly compiled C-applications' (one programe - one model) and flexibility of modern interpreters like GENESIS, NEURON, Python, Java and so on. The recipe was found in "precompiled dynamical library connected to each other by common data". The several inward releases have allowed to crystallize the basic idea and reduce the NeuroCAD Engine code up to 2541 lines with comments (an easiest lightweight kernel). The first stable alpha version was released 15th January 2007 on Lab. NISMS website. Current stable and unstable version can be downloaded here: neurocad.org/download/index.html .

Support

>NeuroCAD project is not supported anymore. (RTH)