Design Automation for Embedded Systems Group

M.Sc. Mikail Yayla

E-Mail: mikail.yaylatu-dortmundde
Phone: +49 231 755-6110
Fax: +49 231 755-6116

Address:

Mikail Yayla
Technische Universität Dortmund
Lehrstuhl Informatik 12
Otto-Hahn-Str. 16
44227 Dortmund
Room E23

Research Interests

Robust and Efficient Machine Learning for Embedded Systems
- Neural Networks
- Decision Trees and Random Forests
Emerging Low-Power Hardware
- Digital and Analog Computing-based Accelerators
- Neuromorphic Accelerators
- Devices with Approximate Components (Memory, MAC-units, etc.)

Reviewer:

Supervised Theses:

Efficient Training of Binarized Neural Networks through Gradient Information Drop (ongoing, Sanad Marji)
Design and Implementation of a Fixed Point FFT Accelerator for Real-time Quantum Spin Noise Spectroscopy (ongoing, Dominik Riemer)
Optimization of Hardware Design for Real-time Quantum Spin Noise Spectroscopy (completed by Sebastian Hilker)
Design and Evaluation of Accelerator Organizations for Binarzed Neural Networks (completed by Somar Iskif)
Efficient online learning in Spiking Neural Networks for drifting data streams (completed by Florian Köhler)
GPU Acceleration for the Inference of Binarized Neural Networks (completed by Leonard Bereholschi)
Design and Evaluation of Superscalar Processor Organizations for Single-Cycle and pipelined Architectures (completed by Robert Huber)
Gradient-based Pruning of Binarized Neural Networks (completed by Ole Dickel)
Hardware Acceleration for Real-time Quantum Spin Noise Spectroscopy (completed by Kevin Goldstein)
Optimization of Processing Elements for Binarized Neural Networks (completed by Umut Yilmaz)