Orange Forum • View topic - GSoC: Statistics Widget, Neural and Others

GSoC: Statistics Widget, Neural and Others

General discussions about Orange and with Orange connected things (data mining, machine learning, bioinformatics...).

GSoC: Statistics Widget, Neural and Others

Postby karthik » Fri Mar 23, 2012 12:57

Hello,
I am Karthik from India - currently pursuing Bachelors in Computer Science and Engineering.
I am very interested in Data Mining and Machine Learning. I have, infact, undertaken courses on Machine Learning and Artificial Intelligence.
I have chosen Data Mining as a special elective from my University and have been using Orange as a tool.

I am particularly interested in the following proposed ideas:
1) Widget for Statistics
--> implementing a few statistical tests as mentioned on the ideas page. I also thought of implementing some performance analysis tools like Calibration plots, detailed ROC analysis, etc.

2) Neural Networks Classifier.
--> My idea is to implement and fully document the Neural Network Classifier. Also, I am hoping to introduce the neural network analyser tools for rules extraction from it - which can help understand the working better.

3) Apart from the two, I also wish to implement some algorithms (with supporting documentation) under other widgets.
--> for example, implementing FP Growth algorithm for Association and so on.

As an Orange user myself, these features are on my personal wish list as well and I would love to work and contribute the same this summer.

Please let me know about your thoughts regarding my ideas and guide me accordingly.

Eagerly waiting for your response.

Thanking you.

Best Regards,
Karthik.

Re: GSoC: Statistics Widget, Neural and Others

Postby jurezb » Tue Apr 03, 2012 13:17

2) Neural Networks Classifier.
--> My idea is to implement and fully document the Neural Network Classifier. Also, I am hoping to introduce the neural network analyser tools for rules extraction from it - which can help understand the working better.


The focus for NNs will probably be on implementing the basic multilayer perceptron. Understaing the inner workings of NNs and presenting it in a set of rules is not that interesting for us at the moment.


Return to General Discussions