With all the hype that Machine Learning is generating recently, we wanted more people to become involved easily. We wanted to create a service through which they could upload any data sets they have, choose what kind of an algorithm - whether they want to find trends in their data, or classify various features of their data - and run it on their dataset without the programming effort.
What it does
ML-Party gives a simple interface to run machine learning algorithms on your datasets, so that you can focus on doing the things that are important and leave the actual machine learning models to us.
How we built it
We created ML-Party from scratch by implementing the math for each algorithm using only Numpy and Scipy's optimization functions.
Challenges we ran into
Learning a lot of these tools and the theory behind them and implementing them over this weekend was a fairly complicated and time consuming task, but it was extremely rewardig too. Also at multiple points, our hosted code was not performing optimally when supposed to, and that was a tough task to deal with.
Accomplishments that we're proud of
We actually got a number of different algorithms working with a consistent efficiency and simplistic interface. Considering that neither of us had much experience with machine learning theory before the event, our finished tasks were extremely rewarding.
What we learned
The theory behind a number of different machine learning algorithms, such as KNN, Naive Bayes, Linear Regression, Logistic regression and Neural Networks.
What's next for ML-Party
Writing the code for SVM and random forest classifiers, and finally launching the project as a complete python library.