-Machine Learning – Prof.Mikhail Belkin – Spring 2017
Yelp recently has initiated a Data mining / Machine learning competition and invited students to explore Yelps dataset and discover interest-ing insights and patterns. The Interaction between the user and the Yelp application is based on the reviews
which are initiated some keywords potentially ranked by ratings. The goal of this project is to explore the reviews from users on Yelp.com and answer if a review is positive or negative. Reviews are stored as plain text and requires some Natural Language Processing in order to give it a structure and transforms it into a machine-learning usable format.
– Introduction to High Performance Deep Learning – Prof.Dhabaleswar K. (DK) Panda – Fall 2018
Evaluation of Error Propagation in Deep Learning Neural Network (DNN) Applications and Accelerators : The goal of this project is to understand soft error vulnerability of DNN system and enable us to propose solution to handle soft error occurred in DNN hardware accelerator.
– High Performance GPU Computing – Prof. P.(Saday)Sadayappan & Dr.Aravind S. Rajam -Fall 2018
Bitmap indexes, supported by fast bitwise operation, are a widely used structure
in the database area from query optimization to approximate aggregate queries. I design Parallel Bitwise Operation toolkid in CUDA without uncompressing the bitvector.
– Senior Graduate Project Prof.Hakan Ferhatosmanoglu( Bilkent University ) – 2015
Whizper is a location based anonymous social platform which contains dynamically created “cluster”s across the globe. If user is within the range of a cluster, (s)he can post a content in that cluster and receive comments from (optionally) anonymous people (who are also in that cluster). If user is not within the range of a cluster, (s)he can view posts in the cluster but (s)he can not contribute to that cluster. We applied clustering and data mining technique for this project.
– Programming Language – Prof.Neelam Soundarajan – Spring 2017
A Lisp Interpreter written in C++
– Computer Architecture – Prof.Radu Teodorescu – Fall 2017
Branch Prediction : Various branch prediction implemented in class project (Brach Prediction Championship). The predictor was evolved from PPM->TAGE->L-TAGE. Idea is taken from the paper titled “A 256 Kbit L-Tage branch predictor”
