| CmpE WEB

Ontology Based Categorization of Habitat Entities using Information Retrieval Techniques

Gathering information about habitats of bacteria and representing them in a normalized way is the key point of microbiology studies. Having this information spread through textual resources such as
scientific articles and web pages leads to a need for automatically detecting habitat entities in text, semantically tagging them using ontologies, and finally extracting the events among them. These

A Hashtag Segmentation System for Turkish

Twitter is one of the most popular social network and microblogging tools worldwide.It is free to use and has over 300 million users totally, and estimated turkish users are over 10 million.Users can send and read messages with limit of 140 characters.These messages are called “tweets”.Tweets can be retweeted by other users or can be saved through using “Like” button.Tweets appear on user’s profile page.On Twitter and many other social microblogging applications, contents may include hashtags to group and categorize them.However hashtags can consist of a word or a combination of words.In th

Parkinson Monitoring by using Mobile Systems

In the project Mobile Application for Parkinson Monitoring, the very beginning starting point for us was exhibiting a product that helps medical personnel and doctors to track and interpret the symptoms of a Parkinson patient better compared to old-fashioned order. For this purpose we did use EXLs3 branded inertial measurement units and a mobile application to control, parse and analyze the daily movement and trembling data of patients.

Decision Tree Learning Algorithms

We worked with different algorithms: C4.5, Linear Support Vector Machines, Multilayer Perceptron and most importantly Soft Decision Trees and Budding Trees. Last two were proposed by our Advisor Prof. Ethem Alpaydın and his former student Ozan İrsoy. We implemented all of them and tested on different large data sets to see the potential of those trees. We tried to understand the learning process through different methods. Our aim was to spot weak points of those algorithms and propose new models based on those. We list many results and visuals on large data sets.

Learning Sequences Using Recurrent Neural Networks

Songs, poems, books, movies, dialogues are just some examples of data found in a sequential form. Learning these sequences, more importantly, being able to generate these sequences is a difficult task. Recurrent neural network is a model which can create a dynamic internal state that withholds information about the data. A Long Short-Term Memory neural network is a more advanced architecture that can remember its ’experience’ for a very long time sequence, thus better adapt to the input. This research in-depth analyzes the structure of the LSTM neural network.

Row Store and Column Store Comparison in Clinical Data

Clinical data is growing repeatedly with its characteristics such as sparseness, high dimensionality and changing set of attributes. Thanks to the surveys hold by Search Health IT, we know that medical center administrators spend more money for health care analytic tools in every year and they will keep spending. They want to analyze the data with better performance. We aimed to illuminate them by comparing the row store databases and column store databases. Also we showed that how the performance would be if there is a hybrid structure.

Implementation Of Adaptive Indexing On Environmental Data

This project is about implementing adaptive indexing on environmental data. The goals of this project is understanding of database indexing, how to implement different indexing methods to databases, how query retrieval times vary in different methods and sizes.

Implementation of a Realtime Pitch Tracker

This project is about implementing a VST plugin that works on a host interface. “The goals of this project are learning the C VST SDK MIDI and Audio Programming implementation of algorithms for real time pitch tracking”. A VST(virtual studio plugin) is a plugin that integrates software audio synthesizer and effect plugins using DSP(digital signal processing). A VST works on a DAW(digital audio workstation).

Implementation of a Realtime Tempo Tracker as a VST Plugin

This project aims at implementing a VST plugin that tracks the tempo of a musical piece in BPM (beats per minute) in real time, namely the RealTime Tempo Tracker plugin. The implementation of the RealTime Tempo Tracker plugin comes in two parts: onset detection and measuring the tempo using a Hidden Markov model based on onset data. The plugin takes some waveform as input via a microphone or a stereo mixer. Then Fast Fourier transform is applied on windows of the waveform to track the change of the power spectrum over time.

Logistic Regression for Sales Prediction

In this paper, we present a system to predict sales of a product. It is based on logistic regression technique. We train the algorithm using the past sales data of a certain product and we predict whether a customer will buy the product in the following day. We experiment with various parameters, such as regularization constant and number of customers and show the results.

Search form

Main Menu