Wrapper around daft for easily creating DBN.

Install

Alternative A) Install systemwide

python setup.py build
sudo python setup.py install

Alternative B) Install dependencies + include

pip install daft --user

1. Usage

see examples

Example: Creating a Model

This example creates a HMM with two explicitely visualized parameters pi and Sigma:

dbn = DBN()
dbn.attach(NodeProperties("X", 0, 0, parentsPrevious="X"))
dbn.attach(NodeProperties(name="Y",x=0, y=1,parentsNow="X", nodeType=NodeType.Observed, continuous=True))
dbn.attach(NodeProperties(name="\Sigma",x=0, y=2,parentsNow="Y", nodeType=NodeType.Variable))
dbn.attach(NodeProperties(name="\pi",x=0, y=2,parentsPrevious="Y", nodeType=NodeType.Variable))

Example: Exporting a Model

Depending on whether the first few time slices are to be visualized (1) or the time slices around the current time, the plot can be exported via

k.export(sliceBefore=0, sliceAfter=3, centerSuffix="") #< variant 1:  export first few time slices

or (2):

k.export(sliceBefore=2, sliceAfter=1, centerSuffix="\\tau") #< variant 2: export 2 before - 1
                                                            #  after the current time (referred
                                                            #  to as \tau)

These 7 lines of code create the following two vector graphics:

hmm1.pdf and hmm.pdf

Output - format:

Supports pdf svg jpg and a lot more.