Thesis supervisor: János Botzheim
Location of studies (in Hungarian): Helyszín angolul Eötvös Loránd University, Faculty of Informatics Abbreviation of location of studies: IK
Description of the research topic:
Several requirements can be set against applied AI systems. Among these the most common requirements target the performance, or accuracy of the given AI application. However, as autonomous mobile systems begin to show up, other requirements are gaining importance, such as robustness and energy efficiency. This research topic focuses on low energy consumption neural networks, more specifically on the third generation of neural networks, spiking neural networks.
The tools related to the field of such energy efficient networks differ from those applied in mainstream Deep Learning, the reimplementation of Deep Learning techniques and algorithms is usually not a trivial task. The task of the participating student is to carry out basic research and analysis targeting the architectural improvements of these neural networks, while maintaining their property of low energy consumption. Besides state-of-the-art Deep Learning techniques, research directions build on the utilization of other fields such as computational intelligence, mathematics (swarm intelligence, neuroevolution, fractional calculus), and human neurobiology, thus advantageous properties of the human nervous system may be implemented in artificial spiking neural networks as well. Research directions target architectural novelty accompanied by the analysis of additional features, like scale-free operation, robustness, online trainability, and event-driven behavior.
Further aims of the dissertation are to examine research results in the light of automotive applications (such as AI assisted sensor data analysis, ADAS systems, and self-driving vehicles), and to investigate the hardware implementation (on neuromorphic processors or FPGAs) of spiking neural networks.
Required language skills: English Recommended language skills (in Hungarian): Hungarian Further requirements: SW: python (Numpy, Pandas, Matplotlib), machine learning (TensorFlow, scikit-learn, Pytorch, Keras)
Field related knowledge: Mathematics of Deep Learning, basics of spiking neural networks
Proactivity, good communication skills
Deadline for application: 2023-05-31
2024. IV. 17. ODT ülés Az ODT következő ülésére 2024. június 14-én, pénteken 10.00 órakor kerül sor a Semmelweis Egyetem Szenátusi termében (Bp. Üllői út 26. I. emelet).
Login
