GMN Methods

GMN

Description :
Class object for Generative Manifold Networks (GMN).

Constructor Signature

class GMN:
    def __init__( self,
                  args        = None,  parameters = None,
                  configFile  = None,  configDir  = None,
                  outputFile  = None,  cores      = 4,
                  backend     = None,  chunks     = None,
                  kdWorkers   = None,  kernel     = None,
                  plot        = False, plotType   = 'state',
                  plotColumns = [],    plotFile   = None,
                  figureSize  = [8,8], verbose    = False, debug = False ):

Full configuration is performed by instantiating GMN with a config file: gmn = GMN( configFile = 'myConfig.cfg' ). It can also be configured passing in args from gmn.CLI_Parser.ParseCmdLine() and parameters from gmn.ConfigParser.ReadConfig(). This is done by the command-line-interface (CLI) program Run.py. RunNoConfig.py does neither: it has its own argument parser and populates a Parameters object directly from the command line, since its purpose is to run without a config file.

If args is None GMN.__init__ function arguments are used to populate args.

If parameters is None the Parameters object is read from the config file named by args.configFile, via ConfigParser.ReadConfig(). Command line values are not merged into it: they remain on args.

A config file may specify any subset of the recognized keys. Keys omitted from the file keep the default assigned in Parameters.__init__.

backend and kernel default to None so that a config file can set them. Each is resolved as: an explicit constructor or CLI value, otherwise the config file [GMN] key, otherwise the Parameters default ('serial', True).

Parameter	Type	Default	Purpose
args	argparse ArgumentParser	None	command line arguments
parameters	gmn Parameter object	None	GMN parameters
configFile	string	None	Configuration file for Network / Nodes
configDir	string	None	Path to directory of configuration file(s)
outputFile	string	None	Generated data output file : .csv or .feather
cores	int	4	Number of CPU processor cores
backend	string	None	'serial' or 'parallel'
chunks	int	None	parallel chunk size
kdWorkers	int	None	thread workers in kdTree.query() : pyEDM only
kernel	bool	None	True - float32 Simplex kernel : False - pyEDM node functions
plot	bool	False	Logical to plot time series results
plotType	string	'state'	'time' or 'state' plot
plotColumns	list	[]	List of columns to plot
plotFile	string	None	File for plot results
figureSize	list	[8,8]	Plot figure size in inches
verbose	bool	False	Logical for verbose output
debug	bool	False	Logical for debug output, enables faulthandler

Returns :
GMN class object. The object is initialized to create the GMN.Network class object, all Node class objects of the network including input data, and the GMN.DataOut pandas DataFrame.

Example :

# Initalize GMN object with default.cfg
import gmn
G = gmn.GMN( configFile = 'config/default.cfg' )

---

GMN.Generate

Description :
Execute GMN forecast loop for predictionLength steps calling the Generate() method of each Network Node. Parameters mode must be Generate.

Returns :
Populates the GMN.DataOut pandas DataFrame.

Notes : If args.outputFile, or parameters.dataOutFile: write DataOut as a .csv or .feather file according to the dataOutFile file extension.

if args.Plot or args.StatePlot or parameters.showPlot or parameters.plotFile: call GMN.Plot()

Example :

G.Generate()

---

GMN.Forecast

Description :
Execute GMN Forecast() method of each Network Node. Parameters mode must not be Generate. Presumes Parameters lib and pred are specified in the config file. Does not generate data, but makes predictions over the pred indices based on the lib state-space.

Returns :
Populates the GMN.DataOut pandas DataFrame.

Notes :
If args.outputFile, or parameters.dataOutFile: write DataOut as a .csv or .feather file according to the dataOutFile file extension.

if args.Plot or args.StatePlot or parameters.showPlot or parameters.plotFile: call GMN.Plot()

Example :

G.Forecast()

---

GMN.Plot

Description :
Plot generated time series (args.Plot = True, or Parameters.plotType is 'time') or time series and 2-D state-space plots (args.StatePlot = True, or Parameters.plotType is 'state').

Returns :
pyplot image

---

GMN Attributes

GMN.DataOut

Description :
pandas DataFrame of generated data.

GMN.Parameters

Description :
Python object of Parameters class.