Hardware Accelerator for Deep Learning

Back to all technologies
Download as PDF
2017-CULU-67658
Deep neural networks are mathematical models designed to perform classification and localization of objects within images and videos. They are also used for speech recognition, generating text from images and several other computer vision applications. These models are computationally very expensive and generally require the use of hardware that is extremely energy intensive.

Researchers at Purdue University have designed a hardware accelerator for deep learning along with code written to implement it. The technology described, when implemented on field programmable gate array devices, is designed provides performance while requiring a fracture of the energy.

Advantages:
- Power efficient
- Improves computational efficiency

Potential Applications:
- Neural Networks
- Speech Recognition\
May 25, 2018
Utility Patent
United States
(None)
(None)

May 26, 2017
Provisional-Patent
United States
(None)
(None)
Purdue Office of Technology Commercialization
1801 Newman Road
West Lafayette, IN 47906

Phone: (765) 588-3475
Fax: (765) 463-3486
Email: otcip@prf.org