Here will be solutions to Programming Exercises proposed during Stanford's online Machine Learning class. Some notes, thoughts and reasoning on course might be also added.
Contributions are highly appreciated.
If you are attending the class, feel free to fork the repo, open issues, request pulls, etc.
Paying respect to the Honor code, I will not publish 100%-correct solutions before due time, such solutions will be added shortly after the deadline. However, work-in-progress solutions will be available.
More information about this course you could find on official website and in wiki.
For information about Octave, download links, installation instruction, etc. go to GNU/Octave or wiki.
Solution to second exercise, Logistic Regression, have been added.
Rough draft of digit recognition (Ex 3).
Blank for 4th exercise, Neural network learning, has been added.
Some attempts on Ex 4.
Some math and logic on why to choose epsilon = 0.12 for Weights Initialization could be found here in section 4.2.1 «Theoretical Considerations and a New Normalized Initialization».