Structured weight matrices-Based hardware accelerators in deep neural networks: FPGAs and ASICs

Engineering & Materials Science

• Application specific integrated circuits 100%
• Computational complexity 19%
• Computer hardware 55%
• Convolutional neural networks 21%
• Deep neural networks 91%
• Energy efficiency 55%
• Field programmable gate arrays (FPGA) 77%
• Industry 10%
• Integrated circuit design 28%
• Internet of things 19%
• Long short-term memory 54%
• Particle accelerators 95%
• Throughput 16%