Yi Zhang

Gateway Programming for ADAS (BMW)

Topic: Genetic Algorithm-based Neural Network Compression for Neural Signal Processor - Implementation and training of a neural network for panoptic segmentation - Connection of an embedded controller to the genetic search algorithm - Design of a novel method for pruning the neural network for panoptic segmentation using the multi-objective genetic search algorithm and the embedded control device Applied programming languages: Python in combination with the deep learning frameworks Tensorflow/Keras and ONNX

- Integrating the NVDLA accelerator on the target FPGA Xilinx Zynq Ultrascale+ MPSoC with Vivado - Generating customized system images with Petalinux for the FPGA - Exploring the challenges of accelerating the object detection algorithm Tiny YOLO on the NVDLA accelerator - Bit-flip resilience analysis of the object detection model Tiny YOLO Applied programming languages: Python in combination with the deep learning framework Caffe and the inference acceleration framework TensorRT; C++

- Programming of scripts for data acquisition and analysis of test database to optimize control unit inventory evaluation - Further development of a tool for checking the quality of test scripts - Processing of ECU data - Help with the development of the software quality checker Applied programming languages: Python 2 and Python 3 including the GUI framework wxpython

- Expansion of a finite-difference time-domain solver for 1D Maxwell-Liouville-von Neumann equations used in quantum cascade laser simulation by adding support for two relevant boundary conditions - perfectly matched layer and partially reflecting mirror - Simulation and verification of the principle of both boundary conditions in Matlab - Implementation of both boundary conditions in the solver written in C++ Applied programming languages: C++, Matlab

Course: practical part of the lecture "Stochastische Signale" Helping students solve homework problems in Matlab in the area of probability theory and signal processing

Grade: 1.3 Master Thesis: Genetic Algorithm-based Neural Network Compression for Neural Signal Processor (Grade: 1.0)

Grade: 1.8 Bachelor Thesis: Extension of the mbsolve Simulation Software: Implementation of Flexible Boundary Conditions (Grade: 1.3)