AdaGrad classification version added; Data set from CS 446 HW 3 added

[yj14n9xyz]

@danyaljj @christos-c Please review this PR.

• AdaGradCL classification added
• AlgoDataSet: CS 446 Homework 3 data set
• AlgoData, AlgoParser: example data structure, parser
• AdaGradCLClassifier added

With data set of 500 features and 50000 examples (40000 for training, 10000 for testing), AdaGradCL has 99.61% accuracy with only 10 iterations.

Thanks!

[danyaljj]

Do you think there is a way to have a single AdaGrad for both regression and classification?

[yj14n9xyz]

@danyaljj That is not quite possible, given the current inheritance architecture.

They both share some same interface, from Learner, such as classify. And, the necessary components for regression and classification are dramatically different, as the implementation shows. Classification and regression have their own unique set of functions to work properly.

[yj14n9xyz]

@danyaljj FYI:

scikit-learn, a popular python ML lib, provides SGDClassifier and SGDRegressor.

[yj14n9xyz]

@danyaljj @christos-c Made a minor update. Added a debug utility class to help with debugging. I found it quite useful.

Please review the changes.

[christos-c]

@yimingjiang what is the status of this? Do you want to update it?

[yj14n9xyz]

@christos-c I will make a new PR soon. This PR is outdated. I will close it.