Fall

Tool to determine microservice boundaries for software-defined vehicles

This project aims to develop a tool that analyzes the functionality and dependencies of different software components in a vehicle and recommends the optimal boundaries for dividing the system into microservices that follow the publisher-subscriber pattern. The goal of this project is to improve the scalability, flexibility, and maintainability of software-defined vehicles by breaking down monolithic systems into smaller, modular, and independently deployable services that can communicate using the publisher-subscriber architecture.

Turkish Preprocessing Operations Using Deep Learning Approaches

The first step in nearly all natural language processing (NLP) applications is applying preprocessing operations [1] to the text. Preprocessing operations include tokenization (segmenting the text into tokens), sentence splitting (dividing the text into sentences), normalization (converting the text into a canonical form), and the like. In this project, you will develop and implement algorithms for preprocessing of Turkish text using deep learning approaches. First, a literature review will be conducted and similar systems for English will be analyzed (e.g. UDPipe [2], Stanza [3]).

Building a word embeddings repository for Turkish

In this project, we aim at building a comprehensive word embedding [1] repository for the Turkish language. Using each of the state-of-the-art word embedding methods, embeddings of all the words in the language will be formed using a corpus. First, the three commonly-used embedding methods (Word2Vec [2,3], Glove [4], Fasttext [5]) will be used and an embedding dictionary for each one will be formed. Then we will continue with context-dependent embedding methods such as BERT [6] and Elmo [7].

Social Network Analysis on Research SONAR invited you to Social Network Analysis on Research

Digitalizing Weather Records on Kandilli Rasat Notebooks

Kandilli Meteorology Laboratory has been making continuous climate measurements at the same point since July 1911. Kandilli Meteorology Laboratory maintains its feature of being the longest climate data of Turkey with 111 years of data formed as a result of climate measurements. We want to start a study to convert this data into digital form.

Here is the data we have:

RF Monitoring via RTL-SDR

Berief Description:

Monitoring RF environment has varios applications, one of them can be detecting communication devices for security purposes.

Social Media App for Low-Bandwidth Distributed WiFi Networks

Berief Description:

The scope of this project includes Wireless Communication under disaster scenario and a chat application for information dissemination feeds in a distributed manner.

Analysis of Model Poisoning Attacks during training in Federated Learning

Federated learning is a distributed machine learning technique that aggregates every client model on the server side. There can be various types of attacks to destroy the robustness of this learning system. Model poisoning attack is realized after the training is finished. In this project, we will use a recently developed method* for analyzing the effect of model poisoning attack that might occur during training.

Model Poisoning Attack Detection in Federated Learning

Federated learning is a distributed machine learning technique that aggregates every client model on the server side. There can be various types of attacks to destroy the robustness of this learning system. A recent study* introduces a low-cost approach for the server to detect these malicious models by coordinate-based statistical comparison. In this project, we will extend this method for detecting model poisoning attacks both on the clients and on the server.

Multi-label classification for concurrent activities using wearable-sensor data

We frequently perform multiple activities simultaneously in daily life, such as drinking coffee while walking or conversing with a friend while standing. When the traditional single-label output is preferred for activity recognition with wearable sensor data, this requires multiple classes with many possibilities for concurrent activities, such as creating a class for "drinking coffee while walking" and another class for "talking while walking". Instead, this study will investigate a multi-label output system to provide information about which activities are performed together.

Search form

Main Menu