Physical Processes and Model Variables
The Neurospaces Description Format or NDF file format is a declarative ASCII text file format indicated by the .ndf file name extension. NDF files are purely declarative in nature. Rather than giving direct access to procedural language constructs, NDF files integrate with a library of procedural algorithms. Each algorithm in this library is a procedural abstraction that has full access to the model, can query it, add new components, and delete or modify existing components. Alternatively an algorithm can analyse the model structure and report the result to the user.
At compilation time the library of algorithms is integrated with the model-container. Each algorithm has a unique name that can be used from inside an NDF file to run the algorithm. Some algorithms may require the assignment of specific parameter values to determine a deterministic outcome. This assignment of parameter values is given in the NDF file where the algorithm is referenced. After the algorithm has completed, the model-container stores the parameter assignment and additional specific algorithm results as an algorithm instance. Such an algorithm instance can then be retrieved by the user by name.
The following examples runs the algorithm with name ’Grid3D’ and assigns the name ’GranulePopulation’ to the algorithm instance. The outcome of this NDF snippet will be 6240 neurons instantiated from the prototype “Granule_cell” and laid out in 3D space.
PARAMETER ( PROTOTYPE = "Granule_cell" ),
PARAMETER ( X_COUNT = 120.0 ),
PARAMETER ( X_DISTANCE = 2.5e-05 ),
PARAMETER ( Y_COUNT = 26.0 ),
PARAMETER ( Y_DISTANCE = 1.875e-05 ),
PARAMETER ( Z_COUNT = 2.0 ),
PARAMETER ( Z_DISTANCE = 2e-5 ),
The model-container algorithm library complements the NDF file format with procedural descriptions. Algorithms in this library include Grid3D (to define a 3D grid of other components), Spines (to add spines to a morphology by calculating surface area corrections or by attaching them given a prototype spine), ProjectionVolume and ProjectionRandomized (to create projections and connections between two populations of neurons) and Randomize (to randomize certain parameters of a named model)