Brandenburg / Bayern Action for AI-Hardware (BB-KI Chips)

Project Description

This project is motivated by the mismatch between the deficiency of hardware-related content in AI curricula of universities and the fast iteration of dedicated hardware optimized for AI tasks by industry. In this context, BB-KI Chips aims to provide an organized curriculum conformed to the 4C/ID model, containing a range of courses covering topics including hardware/chip design, computer architecture, embedded development, machine learning, FPGA, etc., supported by a cross-university (Universität Potsdam and Technische Universität München), multidisciplinary group for diverse application scenarios. In our chair, we focus on efficient edge AI computing on FPGAs. Concretely, given a trained deep learning model for deployment in edge environments, we try to compress the model by pruning and quantization and utilize the massive scale parallelism and dedicated hardware optimization empowered by Xilinx® FPGAs with Deep Learning Processing Units (DPUs) to achieve fast AI inferences.

Project Results

The project is at the moment reaching the second teaching phase in which we design and implement elements of the teaching curriculum for both an AI-aware chip design education path as well as for a chip design-aware AI education path.

Project outcomes including first versions of complete curricula will be published here and on the web page of the project consortium: https://www.bb-ki.de/.

Publications

1. Xiong, Z., Stober, D., Krstić, M., Korup, O., Arango, M. I., Li, H., & Werner, M. (2023). Integrating AI Hardware in Academic Teaching: Experiences and Scope from Brandenburg and Bavaria. ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences. [BibTeX]

   @article{2022_IGCE_AcademicTeaching_BBKI_Xiong,
     title = {Integrating AI Hardware in Academic Teaching: Experiences and Scope from Brandenburg and Bavaria},
     project = {bbki},
     author = {Xiong, Zhouyi and Stober, Dirk and Krstić, Miloš and Korup, Oliver and Arango, Maria Isabel and Li, Hao and Werner, Martin},
     journal = {ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences},
     year = {2023}
   }
   

2. Teuscher, B., Xiong, Z., & Werner, M. (2023). An Augmentation Framework for Efficiently Extracting Open Educational Resources from Slideshows. 9th International Conference on Higher Education Advances (HEAd’23). https://doi.org/10.4995/HEAd23.2023.16169 [PDF] [Online] [BibTeX]

   @article{2023_AugmentationFrameworkOER_Teuscher,
     title = {An Augmentation Framework for Efficiently Extracting Open Educational Resources from Slideshows},
     journal = {9th International Conference on Higher Education Advances (HEAd’23)},
     author = {Teuscher, Balthasar and Xiong, Zhouyi and Werner, Martin},
     year = {2023},
     url = {http://dx.doi.org/10.4995/HEAd23.2023.16169},
     doi = {10.4995/HEAd23.2023.16169},
     project = {bbki}
   }
   

Acknowledgement

This project is funded by the BMBF - Bundesministerium für Bildung und Forschung (2021-2025)