On Wed, Mar 15, 2017 at 11:13:12PM +0530, abhinav kannan wrote: > Greetings, developers! > > First and foremost, hearty congratulations to the mlpack team for being > accepted in GSoC '17. I am honored to start an interaction with this > hardworking and enthusiastic group. > > I am Abhinav Kannan, a second year student of Computer Science at SRM > University, Chennai, India. I am skilled at C++ and Python, with over three > years of experience in the former, and am a lover of machine learning. > Also, when it comes to computer studies, I am a very enthusiastic, fast and > committed learner, having a knack of picking up concepts and solving > problems by myself. > > I have nearly completed Andrew Ng's machine learning course by Stanford > University on Coursera (certificate due in two weeks!), and since the > course has been a worthwhile challenge, I am looking to take up an > assignment that is bigger and tougher, yet achievable, over the summer. > The project "Parallel stochastic optimization methods" is one which I am > motivated to begin work on. I am currently in the installation phase of > mlpack, running Ubuntu. Since it has been taking some time working by > myslef around the installation, I've decided to reach out here before it's > too late, as in the meanwhile, I have done some background research on the > topics related to the project, which are new to me, such as SCD. Now having > a basic idea of it, I did notice the limitations as well in this technique > here <https://en.wikipedia.org/wiki/Coordinate_descent#Limitations>[1] and > here > <http://stats.stackexchange.com/questions/146317/coordinate-vs-gradient-descent> > [2]. > > With respect to this task, Prof Ng's course gives an in-depth understanding > of optimization algorithms such as gradient descent, alongside linear > regression and logistic regression, with convex functions. Also covered in > the course are stochastic (and batch) gradient descent, neural networks, > map-reduce and data parallelism. There have been tests on these topics in > the course's programming assignments and quizzes, which I have completed, > and hence am confident of my understanding in these topics. > > I am also getting started with multi-threaded programming, as part of my > coursework at university. I would love to begin work on this project soon > (if not right away!), along with some inputs from the developers here. I > feel contributing to mlpack will not just be an honor, but also give me a > challenging, hands-on and memorable experience, besides enhancing my > knowledge and helping me be part of the open source community. > > Looking forward to an early response.
Hi Abhinav, Thanks for getting in touch. It sounds like you have already done a good amount of searching, but just in case, here are two useful pages: http://www.mlpack.org/gsoc.html http://www.mlpack.org/involved.html Those could be helpful for making your first contributions to the library. For the parallel stochastic optimization methods project, you may want to take a look at the mailing list archives. Here's one post about it that describes how open-ended the project is: https://mailman.cc.gatech.edu/pipermail/mlpack/2016-March/000752.html Thanks, Ryan -- Ryan Curtin | "Sometimes, I doubt your commitment to Sparkle [email protected] | Motion!" - Kitty Farmer _______________________________________________ mlpack mailing list [email protected] http://knife.lugatgt.org/cgi-bin/mailman/listinfo/mlpack
