US Patent Application 18218426. SEMI-PROGRAMMABLE AND RECONFIGURABLE CO-ACCELERATOR FOR A DEEP NEURAL NETWORK WITH NORMALIZATION OR NON-LINEARITY simplified abstract

SEMI-PROGRAMMABLE AND RECONFIGURABLE CO-ACCELERATOR FOR A DEEP NEURAL NETWORK WITH NORMALIZATION OR NON-LINEARITY

Organization Name

Microsoft Technology Licensing, LLC

Inventor(s)

Stephen Sangho Youn of Bellevue WA (US)

Steven Karl Reinhardt of Vancouver WA (US)

Jeremy Halden Fowers of Seattle WA (US)

Lok Chand Koppaka of Bellevue WA (US)

Kalin Ovtcharov of Snoqualmie WA (US)

SEMI-PROGRAMMABLE AND RECONFIGURABLE CO-ACCELERATOR FOR A DEEP NEURAL NETWORK WITH NORMALIZATION OR NON-LINEARITY - A simplified explanation of the abstract

• This abstract for appeared for US patent application number 18218426 Titled 'SEMI-PROGRAMMABLE AND RECONFIGURABLE CO-ACCELERATOR FOR A DEEP NEURAL NETWORK WITH NORMALIZATION OR NON-LINEARITY'

Simplified Explanation

This abstract describes a technology that uses a configurable stacked architecture to accelerate operations or layers of a deep neural network (DNN). The architecture includes a fixed function datapath with micro-execution units that perform various operations for a DNN layer. The datapath can be customized based on the specific DNN or operation being performed.

Original Abstract Submitted

The present disclosure relates to devices for using a configurable stacked architecture for a fixed function datapath with an accelerator for accelerating an operation or a layer of a deep neural network (DNN). The stacked architecture may have a fixed function datapath that includes one or more configurable micro-execution units that execute a series of vector, scalar, reduction, broadcasting, and normalization operations for a DNN layer operation. The fixed function datapath may be customizable based on the DNN or the operation.