Fall

Extractive Text Summarization with Supportive Images

The goal of this research is to investigate effects of images in text summarization. Most of the news contain images in the body part, however text summarization only studies text part of the document. It is shown that image captions and titles of texts can improve ROUGE score on text summarization task. However, images might not have captions or extracting captions from images can be challenging task for some datasets. Recent papers about text summarization study multi-document text summarization.

Country Prediction of Users on Social Media Using Machine Learning

The aim of our project is to predict home country of social media users who do not specify it explicitly. To achieve the goal, we used location information of users who already specified it and approach the problem in a probabilistic manner. In this project, we focused on Twitter users, however the scope may be widened to any other social media platform. The problem can be easily extended to other prediction problems with similar missing information structure.

A Web Application for Retrieval and Browsing of Semantic Topics

In this project, we build a web interface to allow those who are interested and the research community to use SBounTI topic identifier - a product of Ahmet Yildirim's PhD study. SBounTI identifies topics of a set of microblog posts of multiple users. It produces semantic web representations for the identified topics. The topics are expressed using an ontology that defines microblog topics.

Ivy: A programming language targeting Ethereum Virtual Machine

Ivy is a programming language designed for Ethereum Virtual Machine. Ivy enables us to write smart contracts and deploy them directly to an Ethereum blockchain.

5G / Cloud RAN

Mobile devices have become a part of our very daily lives. In the last two decades a lot of thing has been changed in the field of mobile communications. The last and very exciting development is the 5G technology which is better and improved version of the LTE as well as supporting green technology.

Cancer Classification via Machine Learning Based on Genomic Data

The human genome contains valuable information about the person and its hand analysis is cumbersome since it consists of billions of nucleotides. This work computationally analyses the mutations in the human genome which are the main source of differences and diseases. It utilizes ensembling of various supervised machine learning models on a cancer dataset that is obtained from TCGA and labels the genomes with their cancer types at 74% test accuracy. The obtained accuracies outperform the works in the literature such as DeepGene and create promising results for the future study.

SERVE:See-Listen-Plan-Act

Serve project is the implementation of a system, that enables interaction with Baxter by speaking.Baxter acts according to given commands.It understands given commands by speech recognition and act in the environment by using object detection.

Understanding Football From a Tactical Perspective

The goal of this project is to understand the game of football from a tactical perspective.In doing so, we look at a number of factors including skills, positions, performance history, game-play and the like.

Concretely, the following are the research questions addressed in this project:

1.Which are the most important skills possessed by football players?
2.Which are the most important team skills that lead to a greater winning percentage?
3.Can similar group of players be clustered based on their characteristics?

A Tool for Finding Correlations with GPU

Some algorithms like statistical calculations in our project are based on arithmetic operations rather than input/output type of operations. Usually this situation means that one part of the code is not affecting the following code and that they can be seperated. Let's give an example. You are playing tetris and what will happen one milisecond later is mostly depended on what your keyboard input will be. Now consider matrix multiplication. When we are multiplying two matrices, we can divide both matrices to four parts and multiply every pair independently.

Baby Cry Classification

Crying is the first oral communication of babies. Babies express their feelings by crying before they learn how to express their feelings. They express their feelings of hunger, sleepy or pain by crying. Sometimes, however, it's hard to find out why a baby cries. This can certainly be frustrating or frightening for a caregiver, so we designed a baby crying classification system to categorize baby crying in this project. With this project we helped parents and nurses meet the needs of babies.

Search form

Main Menu