Orange Forum • View topic - Neural Network Classifier issue:

Neural Network Classifier issue:

A place to ask questions about methods in Orange and how they are used and other general support.

Neural Network Classifier issue:

Postby justsomeguy » Thu Nov 01, 2012 19:12

i am trying to use the Neural Network Learner from Orange on the digits data set from Kaggle found here http://www.kaggle.com/c/digit-recognizer/data

Code: Select all
import Orange
from Orange.classification.neural import *

digitsdata = Orange.data.Table('train.csv')

nnLearner = Orange.classification.neural.NeuralNetworkLearner()
nnClassifier = nnLearner(digitsdata[2:])


this is the output i'm getting:

---------------------------------------------------------------------------
AttributeError Traceback (most recent call last)
<ipython-input-18-b01a5bbf9ebc> in <module>()
----> 1 nnClassifier = nnLearner(digitsdata[2:])

C:\Python27\lib\site-packages\Orange\classification\neural.pyc in __call__(self, data, weight)
191
192 #converts attribute data
--> 193 X = data.to_numpy()[0]
194
195 #converts multi-target or single-target classes to numpy

AttributeError: 'list' object has no attribute 'to_numpy''


running the latest version of Orange on win7 64, running python 2.7 32bit

anyone know whats going on?

Re: Neural Network Classifier issue:

Postby Ales » Fri Nov 02, 2012 11:01

Slicing operation on a Table instance returns a list of Instances and not an new Table instance. Use
Code: Select all
nnClassifier = nnLearner(Orange.data.Table(digitsdata[2:]))

Re: Neural Network Classifier issue:

Postby justsomeguy » Fri Nov 02, 2012 11:42

thanks for the reply.

now i get:

C:\Python27\lib\site-packages\Orange\classification\neural.pyc in __call__(self, data, weight)
191
192 #converts attribute data
--> 193 X = data.to_numpy()[0]
194
195 #converts multi-target or single-target classes to numpy

MemoryError:


which is the same error i get if i don't slice the data at all :?


Return to Questions & Support



cron