Invention Grant
- Patent Title: Floating gate for neural network inference
-
Application No.: US16285925Application Date: 2019-02-26
-
Publication No.: US10909444B2Publication Date: 2021-02-02
- Inventor: Effendi Leobandung
- Applicant: International Business Machines Corporation
- Applicant Address: US NY Armonk
- Assignee: International Business Machines Corporation
- Current Assignee: International Business Machines Corporation
- Current Assignee Address: US NY Armonk
- Agency: McGinn IP Law Group, PLLC
- Agent L. Jeffrey Kelly, Esq.
- Main IPC: G11C16/26
- IPC: G11C16/26 ; G11C16/04 ; G11C7/00 ; G06N3/04 ; G11C11/54 ; G11C16/20 ; G06N3/063 ; G11C11/56 ; G11C7/16 ; G11C11/16 ; G11C15/04 ; G11C7/10

Abstract:
A computer-implemented method, the method comprising: in an initial setup of weights for a floating gate including rows, columns, and a separate input line: comparing a current weight to a desired weight; performing a feedback to the input line to set a voltage to change the floating gate field effect transistor (FET) threshold voltage (VT) and the current weight; and checking that the current weight is within a predetermined tolerance of the desired weight; and performing a stochastic pulse update on the floating gate based on the checking.
Public/Granted literature
- US20190188556A1 Floating Gate for Neural Network Inference Public/Granted day:2019-06-20
Information query