Machine Learning applications
For the Course Project for the course Learning With Kernels we implemented the graph kernels based upon SVN Vishwanathan's 2010 paper on Graph Kernels . The crux of the project was to compute the similarity of 2 graphs . We implemented the Random Walk Kernel and the Shortest Path Kernel on MATLAB and tested it on protein and enzymes data which can be represented as a graph .
IMDB Sentiment Analysis
For the Course Project for the course Machine Learning Tools , Techniques we looked at the imdb movie review dataset to predict the sentiment of a particular movie review . We employed several techniques from Natural Language Processing for example the Google's Word-to-vec and the basic bag-of-words approach and tf-idf vectorizer . We also looked at the Wordnet which was used by Watson as a Semantic Graph for Natural Language Understanding .We looked at several classifiers from Machine Learning Literature and tried most of them and got the best performance using a tf-idf vectorizer and a Support Vector Machine Classifier .
Game Reinforcement Learning
For the Course Project for the course Functional Programming we found no implementations for the most common Reinforcement Learning algorithms such as QLearn and SARSA in Haskell. We implemented the QLearn and SARSA algorithms and also provided a DSEL (Domain Specification Language) to specify the moves and the rewards of a 2-Player game and the engine provides support to train a player against a greedy opponent . We tested a small 2 - player game known as the Cat and Mouse game on our framework and it gave good results .
Collect PDF Pages
For the Course Project for the course Databases we ended up creating a linux app in the "Quickly" framework.We had implemented a system for automatic indexing of slides for later retieval and combining slides to create a new presentation .Since the system was for slides in the PDF format and only looked at the text hence it could also be used by students to automatically create tags for the pages of the book and then retrieve them to create notes for the exam .For the automatic tagging we had used a (Latent Dirichlet Allocation) LDA and Latent Semantic Indexing(LSI) model for automatic indexing and for the database we had used MongoDB .