Integrate Orange with OpenML.org

[brylie]

The OpenML.org website provides buttons to export flows for Weka, MOA, and RapidMiner. E.g.
http://www.openml.org/t/1940

As an Orange user, how do I import a data-flow from OpenML.org into an Orange data flow graph?

Related issues

• OpenML authentication widget #1264 OpenML Authentication widget
• OpenML Data Set widget #1265 OpenML Data Set widget

[kernc]

We don't support loading Weka, MOA, or RapidMiner projects. Orange has its own workflow scheme (.ows). Incidentally, OpenML doesn't (yet) support Orange. If you care, perhaps you can write them and express your interest? 😃

[brylie]

@kernc good suggestion. There is some middle ground here it seems.

[brylie]

I posed the Orange Data Mining + OpenML question on the OpenML discussion list, and got a response:

We definitely hope to support Orange, as well as other Python libraries. We have a Python interface under development: http://www.openml.org/guide#!python

It currently allows you to download data, run experiments, and upload results. We also have some examples in scikit-learn. What is does not yet support are Orange-specific procedures to export Orange workflows.

As far as I know nobody is currently working on the latter but of course we greatly encourage work in this direction and will help as much as we can. It just takes someone (or a few people) to get this started.

[kernc]

Thanks!

It just takes someone (or a few people) to get this started.

Would you like to work on it?

[brylie]

Well, I am rather new to the entire ML realm, having just started learning Orange in the past couple of weeks. I can certainly try, but may need a someone to break the task into small chunks.

[brylie]

I got basic approval to pursue this task at work, so can dedicate some time given guidance.

[brylie]

One starter idea is to create a UI widget that can be used to enter OpenML credentials.

[brylie]

Also, where can I find instructions describing how to set up a development environment? I am running Linux, with primary experience in JavaScript development.

[kernc]

Actually, these are the up-to-date instructions: https://github.com/biolab/orange3#installing If you seem to be missing any packages, install them from your distribution channels.

[brylie]

I have followed the "how to install for development" instructions, and am browsing the source.

Where would be a good place to define OpenML related widget(s)? How do I define UI elements (e.g. is there a tool such as QtCreator or other approach)?

[brylie]

Basically, this task is a lot of "uphill" for me. I am willing to learn, but have very little experience with Python, PyQt, etc. There seems to be very little documentation for me to lean on.

[brylie]

One example of useful documentation would be the Node-RED platform docs, where there are relatively clear instructions about creating new nodes:
http://nodered.org/docs/creating-nodes/first-node

[rahulroxx]

@brylie Documentation is at http://orange.biolab.si/docs/

How do I define UI elements (e.g. is there a tool such as QtCreator or other approach)?

Please don't use QtCreator or QtDesigner for coding widgets , as it will hard at later point of time when dealing with loading and saving settings in Orange . You can use them to design layout of the widget . but widgets should be coded manually .
You can use functions from *gui.py *
https://github.com/biolab/orange3/blob/master/Orange/widgets/gui.py

[brylie]

I am working through the widget development getting started tutorial, but cannot get the example widget to appear. I have done the following:

1. added an orange-demo folder inside /orange3/Orange/widgets/

2. created the file structure as outlined in the tutorial

   • orangedemo
     • icons
       • DataSamplerA.svg
     • __init__.py
     • OWDataSamplerA.py
   • setup.py

3. Add the following to orange-demo/setup.py:

   from setuptools import setup
   
   setup(name="Demo",
         packages=["orangedemo"],
         package_data={"orangedemo": ["icons/*.svg"]},
         classifiers=["Example :: Invalid"],
         # Declare orangedemo package to contain widgets for the "Demo" category
         entry_points={"orange.widgets": ("Demo = orangedemo")},
         )

4. Add the following to orange-demo/orangedemo/OWDataSamplerA.py:

   import sys
   import numpy
   
   import Orange.data
   from Orange.widgets import widget, gui
   
   class OWDataSamplerA(widget.OWWidget):
       name = "Data Sampler A"
       description = "Randomly selects a subset of instances from the data set"
       icon = "icons/DataSamplerA.svg"
       priority = 10
   
       inputs = [("Data", Orange.data.Table, "set_data")]
       outputs = [("Sampled Data", Orange.data.Table)]
   
       want_main_area = False
   
       def __init__(self):
           super().__init__()
   
           # GUI
           box = gui.widgetBox(self.controlArea, "Info")
           self.infoa = gui.widgetLabel(box, 'No data on input yet, waiting to get something.')
           self.infob = gui.widgetLabel(box, '')
   
       def set_data(self, dataset):
           if dataset is not None:
               self.infoa.setText('%d instances in input data set' % len(dataset))
               indices = numpy.random.permutation(len(dataset))
               indices = indices[:int(numpy.ceil(len(dataset) * 0.1))]
               sample = dataset[indices]
               self.infob.setText('%d sampled instances' % len(sample))
               self.send("Sampled Data", sample)
           else:
               self.infoa.setText('No data on input yet, waiting to get something.')
               self.infob.setText('')
               self.send("Sampled Data", None)

5. run pip install -e . (both with and without sudo)

6. run python -m Orange.canvas

I do see that a Demo.egg-info directory was created when running without sudo.

Where might I be going wrong?

[rahulroxx]

@brylie You can find your custom widget as shown below ,

Custome Widgets are under Demo in my case .

[kernc]

If you run python -m Orange.canvas -l 4, do you get any errors? Your orange.widgets entry point is not a tuple, rather a single value enclosed in braces — possibly missing a comma there.

[brylie]

Ok, thanks @kernc. For what it's worth, that code comes directly from the demo widget tutorial

[brylie]

@kernc given the above mentioned entry_points line:

entry_points={"orange.widgets": ("Demo = orangedemo")}

Should the entry_points line be changed to this?

entry_points=("orange.widgets", ("Demo = orangedemo"))

Or should it look like this:

entry_points={"orange.widgets": ("Demo", "orangedemo")}

[brylie]

After some searching (and following a broken link in the Orange documentation), I found an example entry_points definition. So, changing the code as follows seems to resolve the problem:

entry_points={"orange.widgets": "Demo = orangedemo"}

[ajdapretnar]

@brylie How's it going with the widget development? If there's something we need to improve in the documentation (and I'm sure there is), I'd be glad to hear it. Any contribution to a better documentation is highly appreciated. ⭐

[brylie]

@ajdapretnar thanks for following up. I will keep working on a simple widget for now, perhaps an OpenML authentication widget that has text field(s) for authentication details. The authentication details can then be passed into other OpenML related widgets.

[brylie]

As an aside, there is a related ML project called TPOT. In the TPOT repository, there are two ideas related to TPOT + Orange integration:

• Export TPOT pipelines to Orange file format EpistasisLab/tpot#51
• TPOT module for Orange Data Mining? EpistasisLab/tpot#137

[ajdapretnar]

@brylie Sounds great! When you have a PR ready, submit it and our folks here will review it. We really appreciate all the enhancements to Orange. ❤️ Once you're done with the widget, you are very welcome to write a guest blog for us on how the development went and how the widget can be used for OpenML tasks.

[brylie]

I will open an issue to track the "OpenML Authentication" widget development.

[brylie]

I renamed this issue to indicate that it is a high-level issue. I updated the issue description with a link to the first sup-tasks:

• OpenML authentication widget #1264 OpenML Authentication widget
• OpenML Data Set widget #1265 OpenML Data Set widget

[joaquinvanschoren]

Hi, I just noticed this thread. Great that you want to integrate OpenML and Orange. Let me know if you need any help from the OpenML side. You're also welcome to one of our next hackathons to speed things up :).

[brylie]

@joaquinvanschoren when/where are your upcoming hackathons?

[joaquinvanschoren]

Next one is the last week of February (Feb 27 - Mar 30) in Munich. The one after that will be the Lorentz Center (Leiden) early September. We're planning one at EMBL in Heidelberg in between, but no details yet.

…

On Thu, Dec 1, 2016 at 11:14 AM Brylie Christopher Oxley < ***@***.***> wrote: @joaquinvanschoren <https://github.com/joaquinvanschoren> when/where are your upcoming hackathons? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1184 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABpQVzUdBhfvKuuXm3QcszupkNvpaCm6ks5rDp3sgaJpZM4III5Q> .

[brylie]

Cool, I am in Finland, so might be able to 'hop the Baltic' for a hackathon in Germany. Any plans of visiting, for instance, Helsinki? 😉

[joaquinvanschoren]

You mean me personally? If you want to organize an OpenML Hackathon in Helsinki, that would also be great :).

…

On Fri, Dec 2, 2016 at 9:33 AM Brylie Christopher Oxley < ***@***.***> wrote: Cool, I am in Finland, so might be able to 'hop the Baltic' for a hackathon in Germany. Any plans of visiting, for instance, Helsinki? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1184 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABpQV_kC5f-8AHpskiby-aEFjbU-vUZSks5rD9fggaJpZM4III5Q> .

[janezd]

In the interest of keeping the list of issues at manageable size, I'm closing this one, but keeping #1265. This one is too complicated and general, and nothing substantial happened in 2.5 years, while #1265 is small enough to be doable, and I believe the OpenML repository is big and well-organized, so it's worth investing some time into it eventually.