This projects is regarding high available services hosted in the cloud using different cloud technologies.
The project work includes investigating development and deployment of cloud-native services and increasing their availability, fault tolerance and operating costs and energy.
Automated Machine Learning methods using Reinforcement Learning
My research focuses mainly on:
-Design AutoML related models
-Neural Architecture Search
-Hyperparameters optimization
-Performance and Evaluation optimization
-Adaptive methods
BRAIN2SPEECH: EEG alapú kommunikációs agy-gép interfész deep learning módszerekkel
A beszéd az emberi kommunikáció elsődleges és legfontosabb eszköze. Sokan azonban elvesztették ezt a képességüket betegség vagy egészségkárosodás okán. A kommunikációs agy-gép interfész (BCI) célja, hogy természetes vagy ahhoz közeli kommunikációs csatornát biztosítsanak olyan személyek számára, akik fizikai vagy neurológiai károsodás miatt nem tudnak beszélni. A beszéd valós idejű szintézise közvetlenül a mért idegi aktivitásból (EEG) lehetővé tenné a természetes beszédet, és jelentősen javítaná az életminőségét, különösen a kommunikációban súlyosan korlátozott személyek számára.
A hallgató feladata megismerkedni a BRAIN2SPEECH területtel, majd új típusú neurális hálózat architektúrák (pl. konvolúciós és rekurrens hálózatok) kidolgozása és tanítása több beszélő adataival.
Az önálló munka / diplomaterv a BME Beszédtechnológia és Intelligens Interakciók Laboratóriumában készül. A hallgató feladatának a következőkre kell kiterjednie:
- Tekintse át beszédtechnológiában az electroencephalogram-akusztikum becslés szakirodalmát.
- Vizsgálja meg, milyen típusú neurális hálózatokat alkalmaztak eddig a kommunikációs agy-gép interfész területen!
- Vizsgálja meg különböző neurális hálózat architektúrák (pl. konvolúciós és rekurrens hálózatok, ResNet, SkipNet) alkalmazhatóságát.
- Az elkészült modelleket tesztelje objektív mérőszámokkal és szubjektív teszt keretében!
- Munkáját részletesen dokumentálja!
Computer Vision and Natural Language Processing in machine learning
Computer vision (CV) and Natural Language Processing (NLP) are two main subfields of machine learning, and a lot of research is going on there.
These two subfields overlap together in tasks such as text generation out of image (image2text) or vice-versa (text2image).
A main obstacle in the way of teaching models (supervised learning) which are able to perform such tasks is the lack of labeled data, and a way to overcome this is to follow unsupervised learning approach.
The task of the student(s) is to get familiar with those tasks and try to reproduce available solutions in order to be able to improve them later.
No. of students: 1 - 3
contact email: alshouha@edu.bme.hu
COMPUTER VISION AND NATURAL LANGUAGE PROCESSING IN MACHINE LEARNING
Computer vision (CV) and Natural Language Processing (NLP) are two main subfields of machine learning, and a lot of research is going on there.
These two subfields overlap together in tasks such as text generation out of image (image2text) or vice-versa (text2image).
A main challenge that is facing these models (and ML based models in general) is the explaination of the model's output, e.g.: why a certain object appears in a certain image captioning.
The task of the student(s) is to get familiar with those tasks and try to reproduce available XAI (explainable AI) algorithms in order to utilize them later.
Number of students: 1 - 2.
COMPUTER VISION AND NATURAL LANGUAGE PROCESSING IN MACHINE LEARNING
Computer vision (CV) and Natural Language Processing (NLP) are two main subfields of machine learning, and a lot of research is going on there.
These two subfields overlap together in tasks such as text generation out of image (image2text) or vice-versa (text2image).
A new subfield has emerged, i.e. Story Visualization, with the help of the advancement of GANs and Diffusion models.
The task of the student(s) is to explore Story Visualization topic by investigating and utilizing the state-of-the-art models in the field.
No. of students: 1 - 3 contact email: alshouha@edu.bme.hu
Valódi, hanggal beszélgető robot (virtuális ügynök) kialakítása a cél, melyhez az NVIDIA NeMo/RIVA toolkiteket használjuk. Magyar nyelven elsőként valósulhat meg a projekt. Python programozási ismeretek, mélytanulási alapok előnyt jelentenek.
Detection and Mitigation of Distributed Denial-of-Service (DDoS) Attacks on Software Defined Networking (SDN)
The evolution of information and communication technologies has brought new challenges in managing the Internet. Software-Defined Networking (SDN) aims to provide easily configured and remotely controlled networks based on centralized control. Since SDN will be the next disruption in networking, SDN security has become a hot research topic because of its importance in communication systems. A centralized controller can become a focal point of attack, thus preventing the attack on the controller will be a priority. The whole network will be affected if the attacker gains access to the controller. One of the attacks that affect SDN controller is DDoS attacks. The aim of this project is to explore and evaluate one of the common detection and mitigation techniques. Later, further ideas on how to improve the performance of SDN during such attacks are investigated.
Energy Efficiency Evaluation of 5G Radio Access Networks Architectures
The exponential growth of network traffic and the number of connected devices make energy efficiency an increasingly important concern for future mobile networks.
More specifically, because 5G is being deployed at a time when energy efficiency appears to be a significant matter for network ability to consider and serve societal and environmental issues, it has the potential to play an essential role in assisting industries in achieving sustainability goals.
Many architectures are proposed as candidates for 5G and beyond radio access networks(RAN).
The Candidate task is to model, evaluate and compare the energy efficiency of different RAN architectures
Explainable Deep Learning Models for Text-to-Speech Conversational AI
Conversational AI uses machine learning to develop speech-based apps that allow humans to interact naturally with devices, machines, and computers using audio. Several deep learning models are connected to a pipeline to build a conversational AI application. This project aims to study and refine the TTS part in one Conversational AI toolkit (for example, NVIDIA NeMo or SpeechBrain).