Researchers from Soongsil University (Korea) published “An investigation of efficient convolutional neural networks and hardware acceleration”.
“Over the past decade, deep learning-based representations have demonstrated remarkable performance in academia and industry. The learning ability of convolutional neural networks (CNNs) comes from a combination of various feature extraction layers that make full use of a large amount of data.However, they often require significant computational and memory resources while replacing traditional hand-designed features in existing systems.In this review, to improve efficiency of deep learning research, we focus on three aspects: quantized/binarized models, optimized architectures and resource constrained systems Recent advances in lightweight deep learning models and search algorithms network architecture (NAS) are reviewed, starting with layers simplified and efficient convolution, through architectural redesign and optimization. Furthermore, several practical applications of effective CNN have been investigated using various types of hardware architectures and platforms.
Find the technical sheet here. Published in March 2022.
Ghimire, D.; Kil, D.; Kim, S.-h. A survey of efficient convolutional neural networks and hardware acceleration. Electronics 2022, 11, 945. https://doi.org/10.3390/electronics11060945.
How to optimize a processor
There are at least three architectural layers in the CPU design, each playing an important role.
New uses for AI in chips
ML/DL increases design complexity at the edge, but it also adds new options to improve power and performance.
Why hardware dependent software is so critical
It might not be the most glamorous type of software development, but doing it right is critical to the success of any hardware platform.