[jira] [Created] (SYSTEMML-1013) RuntimeError: maximum recursion depth exceeded, when using python API

["Imran Younus (JIRA)" <ji...@apache.org>]

Imran Younus created SYSTEMML-1013:
--------------------------------------

             Summary: RuntimeError: maximum recursion depth exceeded, when using python API
                 Key: SYSTEMML-1013
                 URL: https://issues.apache.org/jira/browse/SYSTEMML-1013
             Project: SystemML
          Issue Type: Bug
          Components: APIs
         Environment: Linux mint (laptop). Spark 1.6.2
            Reporter: Imran Younus


I'm trying to implement non-negative matrix factorization using python API. Here is the code that I'm running:
{code}
from pyspark.sql import SQLContext
import systemml as sml
import numpy as np

sqlContext = SQLContext(sc)
sml.setSparkContext(sc)

m = 2000
n = 2400
k = 10
V = sml.matrix(np.random.rand(2034, 2398))
W = sml.matrix(np.random.rand(m, k))
H = sml.matrix(np.random.rand(k, n))

max_iters = 500
for i in range(max_iters):
    H = H * (W.transpose().dot(V))/(W.transpose().dot(W.dot(H)) + 1e-9)
    W = W * (V.dot(H.transpose()))/(W.dot(H.dot(H.transpose())) + 1e-9)
    
res = H.toNumPyArray()
{code}

If the number of iterations are smaller, the code runs fine. But when I change the number of iteration to, say, 500, I get the following error:

{code}
---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
<ipython-input-1-28769ccee510> in <module>()
     18     W = W * (V.dot(H.transpose()))/(W.dot(H.dot(H.transpose())) + 1e-9)
     19 
---> 20 res = H.toNumPyArray()

/home/iyounus/git/incubator-systemml/src/main/python/systemml/defmatrix.pyc in toNumPyArray(self)
    455         """
    456         if self.data is None:
--> 457             self.eval()
    458         if isinstance(self.data, DataFrame):
    459             self.data = self.data.toPandas().as_matrix()

/home/iyounus/git/incubator-systemml/src/main/python/systemml/defmatrix.pyc in eval(self, outputDF)
    440         This is a convenience function that calls the global eval method
    441         """
--> 442         eval([self], outputDF=False)
    443 
    444     def toPandas(self):

/home/iyounus/git/incubator-systemml/src/main/python/systemml/defmatrix.pyc in eval(outputs, outputDF, execute)
    300     reset()
    301     outputs = convertOutputsToList(outputs)
--> 302     matrix.script.scriptString = performDFS(outputs, execute)
    303     if not execute:
    304         resetOutputFlag(outputs)

/home/iyounus/git/incubator-systemml/src/main/python/systemml/defmatrix.pyc in performDFS(outputs, execute)
    105     for m in outputs:
    106         m.output = True
--> 107         m._visit(execute=execute)
    108     return ''.join(matrix.dml)
    109 

/home/iyounus/git/incubator-systemml/src/main/python/systemml/defmatrix.pyc in _visit(self, execute)
    507             # Traverse the AST
    508             for m in self.op.inputs:
--> 509                 m._visit(execute=execute)
    510             self.op._visit(execute=execute)
    511         else:

... last 1 frames repeated, from the frame below ...

/home/iyounus/git/incubator-systemml/src/main/python/systemml/defmatrix.pyc in _visit(self, execute)
    507             # Traverse the AST
    508             for m in self.op.inputs:
--> 509                 m._visit(execute=execute)
    510             self.op._visit(execute=execute)
    511         else:

RuntimeError: maximum recursion depth exceeded
{code}



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)