Inh.: Dr. Renate Gorre
Fon: +49 (0)7533 97227
Fax: +49 (0)7533 97228
Series in Microelectronics
edited by Qiuting Huang
Mathieu Maurice Luisier
Lukas Arno Jakob Cavigelli
2019. XVIII, 234 pages. € 64,00.
Deep learning and particularly convolutional neural networks (CNNs) have become the method of choice for most computer vision tasks. The achieved leap in accuracy has dramatically increased the range of possibilities and created a demand for running these compute and memory intensive algorithms on embedded and mobile devices. In this thesis, we evaluate the capabilities of software-programmable hardware, dive into specialized accelerators, and explore the potential of extremely quantized CNNs—all with special consideration to external memory bandwidth, which dominates the overall energy cost. We establish that — including I/O — software-programmable platforms can achieve 10–40 GOp/s/W, our specialized accelerator for fixedpoint CNNs achieves 630 GOp/s/W, binary-weight CNNs can be implemented with up to 5.9 TOp/s/W and very small binarized neural networks implementable with purely combinational logic could be run directly on the sensor with 670 TOp/s/W.
About the Author
Lukas Cavigelli was born in Zurich, Switzerland in 1990. He received the M.Sc. degree in electrical engineering and information technology from ETH Zurich in 2014. He then joined the group of Prof. Dr. Luca Benini at the Integrated Systems Laboratory, ETH Zurich, to pursue a doctorate. Since his graduation in 2019, he has continued his work there as a post-doctoral researcher. He has received the best paper award at the VLSI-SoC’13 and the ICDSC’17 conferences and the Donald O. Pederson best paper award of the IEEE Trans. on CAD in 2019.
Direkt bestellen bei / to order directly from:
Hartung-Gorre Verlag / D-78465 Konstanz / Germany