danbar / fglib Goto Github PK

Hello,

When trying the test case (test.txt) with Gaussian variables, I receive the following error

Traceback (most recent call last):
File "npbp.py", line 31, in
belief = inference.sum_product(fg, x4)
File "/home/jdmartin86/.local/lib/python3.5/site-packages/fglib/inference.py", line 63, in sum_product
return belief_propagation(graph, query_node)
File "/home/jdmartin86/.local/lib/python3.5/site-packages/fglib/inference.py", line 42, in belief_propagation
msg = u.spa(v)
File "/home/jdmartin86/.local/lib/python3.5/site-packages/fglib/nodes.py", line 271, in spa
msg = msg.marginalize(n, normalize=False)
TypeError: marginalize() got an unexpected keyword argument 'normalize'

I just copy your code in max-product algorithm example. But I got
AttributeError: 'NoneType' object has no attribute 'normalize'.

I guess you do not initialize the value when running the max-product algorithm.

Hi~
When running make html, I was told that:
"The 'sphinx-build' command was not found. Make sure you have Sphinx
installed, then set the SPHINXBUILD environment variable to point
to the full path of the 'sphinx-build' executable. Alternatively you
may add the Sphinx directory to PATH."
How to deal with this exception?
Thanks~

like

x1 = nodes.VNode("x1", rv.Discrete([0.8, 0.2]))

Convert class Node to an abstract base class.

fg = graphs.FactorGraph()

x1 = nodes.VNode("x1", rv.Discrete)
x2 = nodes.VNode("x2", rv.Discrete)
x3 = nodes.VNode("x3", rv.Discrete)
x4 = nodes.VNode("x4", rv.Discrete)
x5 = nodes.VNode("x5", rv.Discrete)

dist_fa = [
    [
        [0.1, 0.2],
        [0.1, 0.1]
    ],
    [
        [0.2, 0.05],
        [0.2, 0.05]
    ]
           ]
fa = nodes.FNode("fa", rv.Discrete(dist_fa, x1, x2,x3))

dist_fb = [[0.1, 0.4],
           [0.2, 0.3]]
fb = nodes.FNode("fb", rv.Discrete(dist_fb, x3, x4))

dist_fc = [[0.5, 0.1],
           [0.2, 0.2]]
fc = nodes.FNode("fc", rv.Discrete(dist_fc, x3, x5))

fg.set_nodes([x1, x2, x3, x4, x5])
fg.set_nodes([fa,fb,fc])

fg.set_edge(x1, fa)
fg.set_edge(x2, fa)
fg.set_edge(fa, x3)
fg.set_edge(x4, fb)
fg.set_edge(fb, x3)
fg.set_edge(x3, fc)
fg.set_edge(fc, x5)
fg.set_edge(x2, fb)
fg.set_edge(fb, x3)
fg.set_edge(x2, fc)
fg.set_edge(fc, x4)

belief = inference.max_sum(fg,x5)`

for this problem I get Value assignment
x5 0
x3 1
x1 3
x2 1
x4 0
but all of the variables are binary. SO x1 = 3 does not looks right

I am trying to use the sum-product algorithm on a graph of multivariate Gaussians in information form.
I define the factors in the appropriate dimensions and get the following error:

I am not sure what exactly is not working, it seams like messages are not marginalized properly.
Can you please offer some advice?

Here is my example code:

fg_Static_Filter_example_v2.zip

Dear Author,

I got this error after installing this package. Can you tell me what I need to do?

`TypeError Traceback (most recent call last)
in ()
39 # Perform sum-product algorithm on factor graph
40 # and request belief of variable node x4
---> 41 belief = inference.sum_product(fg, x4)
42
43 # Print belief of variables

~\Downloads\BitBucket\python\FGlib\fglib\inference.py in sum_product(graph, query_node)
61
62 # Sum-Product algorithm is equivalent to Belief Propagation
---> 63 return belief_propagation(graph, query_node)
64
65

~\Downloads\BitBucket\python\FGlib\fglib\inference.py in belief_propagation(graph, query_node)
49
50 # Return marginal distribution
---> 51 return query_node.belief()
52
53

~\Downloads\BitBucket\python\FGlib\fglib\nodes.py in belief(self, normalize)
132
133 # Pick first node
--> 134 n = next(iterator)
135
136 # Product over all incoming messages

TypeError: 'list' object is not an iterator`

It looks like inference is not working for non-tree structures.
For example consider the following simple factor graph with nodes x1, x2, x3 and factors fa, fb, fc.

from fglib import graphs, nodes, inference, rv

# Create factor graph
fg = graphs.FactorGraph()

# Create variable nodes
x1 = nodes.VNode("x1", rv.Discrete)
x2 = nodes.VNode("x2", rv.Discrete)
x3 = nodes.VNode("x3", rv.Discrete)

# Create factor nodes (with joint distributions)
dist_fa = [[1.0, 0.0],
           [0.2, 0.8]]
fa = nodes.FNode("fa", rv.Discrete(dist_fa, x1, x2))

dist_fb = [[1.0, 0.0],
           [0.3, 0.7]]
fb = nodes.FNode("fb", rv.Discrete(dist_fb, x2, x3))

dist_fc = [[1.0, 0.0],
           [0.4, 0.6]]
fc = nodes.FNode("fc", rv.Discrete(dist_fc, x1, x3))

# Add nodes to factor graph
fg.set_nodes([x1, x2, x3])
fg.set_nodes([fa, fb, fc])

# Add edges to factor graph
fg.set_edge(x1, fa)
fg.set_edge(fa, x2)
fg.set_edge(x2, fb)
fg.set_edge(fb, x3)
fg.set_edge(x1, fc)
fg.set_edge(fc, x3)

# Perform sum-product algorithm on factor graph
# and request belief of variable node x3
belief = inference.sum_product(fg, x3)

# Print belief of variables
print("Belief of variable node x3:")
print(belief)

The terminal output is shown below.

Traceback (most recent call last):
  File "fglib_example.py", line 35, in <module>
    belief = inference.sum_product(fg, x3)
  File "/home/bradley/.local/lib/python3.6/site-packages/fglib/inference.py", line 63, in sum_product
    return belief_propagation(graph, query_node)
  File "/home/bradley/.local/lib/python3.6/site-packages/fglib/inference.py", line 42, in belief_propagation
    msg = u.spa(v)
  File "/home/bradley/.local/lib/python3.6/site-packages/fglib/nodes.py", line 270, in spa
    msg *= self.graph[n][self]['object'].get_message(n, self)
  File "/home/bradley/.local/lib/python3.6/site-packages/fglib/rv.py", line 255, in __imul__
    return self.__mul__(other)
  File "/home/bradley/.local/lib/python3.6/site-packages/fglib/rv.py", line 212, in __mul__
    if len(self.dim) < len(other.dim):
AttributeError: 'NoneType' object has no attribute 'dim'