Python wrapper for Netica C API

This page gives some basic information about the OpenEarthTools Python wrapper for the Netica C API.

Information about the Netica C API can be found at http://norsys.com/netica_c_api.htm; downloads at available at http://www.norsys.com/downloads/

The Python wrapper is available in the openearthtools trunk, more specifically at https://svn.oss.deltares.nl/repos/openearthtools/trunk/python/applications/Netica/

Installation

Step 1

Windows

The .zip file with the Netica C API contains the file Netica.dll
This .dll file should be copied to the lib directory of the Python Netica package (python/applications/Netica/lib/)

Mac OS X

The Netica C API for Mac OS X includes a static library only, where a dynamic library is needed for use with the python wrapper. You have to compile it yourself. First run in the src folder of the zip file:

./compile.sh

This results in a NeticaEx.o file in the lib folder. Then go to the lib folder and run:

g++ -shared libnetica.a NeticaEx.o -o libnetica.so

Copy the resulting libnetica.so to the lib directory of the Python Netica package (python/applications/Netica/lib/)

Step 2

On all platforms, run setup to install the python package:

python setup.py develop

This setup.py file is located in python/applications/Netica/setup.py

Example of basic usage

Building a bayesian network

The code below shows how to build a Netica Bayesian network from scratch using Python.
The .cas file to train the network can be downloaded here.

from netica import Netica
import numpy as np

# initialize class
ntc = Netica()

# create new environment
env = ntc.newenv()
# initialize environment
ntc.initenv(env)
# create new net
net_p = ntc.newnet('BNexample', env)

# define nodes
node1 = ntc.newnode('Node1', 3, net_p)
ntc.setnodestatenames(node1, "one, two, three")
node2 = ntc.newnode('Node2', 0, net_p)
ntc.setnodelevels(node2, 4, np.asarray([1, 2, 3, 4, 5], dtype='float64'))

# define links
ntc.addlink(parent=node1, child=node2)

# obtain node list
nl_p = ntc.getnetnodes(net_p)
# train with cas file
ntc.revisecptsbycasefile(filename='BNcases.cas', nl_p=nl_p, updating=0, degree=1)

# compile the net 
ntc.compilenet(net_p)
# enable auto updating
ntc.setautoupdate(net_p)

# save net
ntc.savenet(env, net_p, 'BNexample.dne')