Congratulations to Niklas Pollak for finishing his Bachelor’s Thesis!

Symbolic picture for the article. The link opens the image in a large view.

Congratulations to Niklas Pollak for successfully completing his Bachelor’s Thesis in collaboration with the Chair of Computer Science 3 – Computer Architecture on the topic of “Optimization and Evaluation of a Neural Network on a Heterogeneous Embedded Computer Architecture for Use in Neuromuscular Environments”!

Niklas’s thesis focused on optimizing and deploying an existing neural network on a heterogeneous embedded computer architecture for neuromuscular environments. The objective was to meet specific requirements for real-time prediction of hand movements in a hand prosthesis by optimizing inference time and power consumption on the hardware. Through the use of TensorRT, multiple inference environments were created and measured to assess the impact of optimizations on power consumption and inference time. The results showcased a remarkable reduction in inference time from 92 ms to 9.9 ms. Additionally, an optimum configuration was identified, meeting the inference time requirement while operating in the most energy-efficient manner. These findings demonstrate the feasibility of running the neural network in a real-time system with energy-efficient performance.

Congratulations again, Niklas, on this outstanding achievement, and best of luck in your future career!