ULTRA PIPELINED ACCELERATOR FOR MACHINE LEARNING INFERENCE

Title:

ULTRA PIPELINED ACCELERATOR FOR MACHINE LEARNING INFERENCE

Document Type and Number:

Japanese Patent JP2021068454

Kind Code:

A

Abstract:

To provide a method of pipelining inference of a neural network.SOLUTION: A method of pipelining inference of a neural network comprising an i-th layer, (i+1)-th layer, and (i+2)-th layer is provided. The method comprises: processing a first set of i-th values of the i-th layer to generate (i+1)-th values for the (i+1)-th layer; determining the quantity of the (i+1)-th values sufficient for processing; and, upon the determination, processing the (i+1)-th values to generate an output value for the (i+2)-th layer while concurrently processing a second set of i-th values of the i-th layer, where i represents an integer greater than 0.SELECTED DRAWING: Figure 2a

More Like This:

WO/2024/064457	MEMORY MANAGEMENT FOR MATHEMATICAL OPERATIONS IN COMPUTING SYSTEMS WITH HETEROGENEOUS MEMORY ARCHITECTURES
WO/2022/247194	MULTIPLIER, DATA PROCESSING METHOD, CHIP, COMPUTER DEVICE AND STORAGE MEDIUM
WO/2022/108693	ARTIFICIAL NEURAL NETWORK BYPASS

Inventors:

TITASH RAKSHIT
RYAN HATCHER
MALIK AQEEL ANWAR

Application Number:

JP2020178413A

Publication Date:

April 30, 2021

Filing Date:

October 23, 2020

Export Citation:

Click for automatic bibliography generation Help

Assignee:

SAMSUNG ELECTRONICS CO LTD

International Classes:

G06N3/063; G06F9/345; G06F9/38; G06F17/10; G06F17/16; G06G7/60; G06T7/00

Attorney, Agent or Firm:

Tadashige Ito
Tadahiko Ito
Osamu Miyazaki

Previous Patent: ＳＳＤ並びにファームウェアベースＳＳＤブロックの故障に対する予測及�...

Next Patent: 写真に基づいてユーザの顔を認識して活用する方法およびコンピュータシ�...