Survey of Machine Learning in Python

Orange

I don’t know who wrote Orange’s documentation, but it is laid out in the strangest manner. Everytime that I ask myself a question, there seems to be a link for the docs page that would answer the question, but then the page in question does not. For example, I’ve written out an “Orange file” in the format that they describe, and I want to open it. How you may ask? How about we go to the page titled “Loading and saving data”. That seems dead on. Except it’s not… the only thing on that page is a very detailed explaination of what the format is! And how it’s represented in memory after it’s read in. Reading in data was finally found in the page titled “Data table (Table)”. Helpful.

I also took the time to look at orange because it had been compared with the venerable Weka, in that it had a nice gui. With the gui and the widgets you could quickly explore data before taking what you have learned and write automated programs to process all of the data. But the docs don’t make it entirely obvious as to how to launch the gui… I had to actually go through the install output to see where it had copied any non-library scripts. And then I had to install dev-python/PyQt4-4.9.4 with the webkit use flag, because the package installer hadn’t bothered to check. But finally once all that stuff was in place the gui “orange-canvas”, which is a graphical programming tool, launched without a hitch.

It was insanely easy to just plunk a file widget on the canvas and connect it up with three different data visualization widgets: “Attribute Statistics”, “Mosaic Display”, and “Sieve Diagram”.

My first attempt at putting a “Classification Tree” into the mix, met with poor results:

james@basalt ~ $ python Schema.py 
Traceback (most recent call last):
  File "Schema.py", line 105, in 
    ow = GUIApplication()
  File "Schema.py", line 26, in __init__
    self.createWidget('OWClassificationTree', 'icons/ClassificationTree.png', 'Classification Tree', 1, self.signalManager)
  File "Schema.py", line 51, in createWidget
    widgetSettings = cPickle.loads(self.strSettings[caption])
KeyError: 'Classification Tree'


But I simply added a "Classification Tree Graph" after it, and the gui started after that.  Weird.
-----