Invention Grant
- Patent Title: Optimized neuron circuit, and architecture and method for executing neural networks
-
Application No.: US16064427Application Date: 2016-12-07
-
Publication No.: US11222254B2Publication Date: 2022-01-11
- Inventor: Alexandre Carbon , Olivier Bichler , Marc Duranton , Jean-Marc Philippe
- Applicant: COMMISSARIAT A L'ÉNERGIE ATOMIQUE ET AUX ÉNERGIES ALTERNATIVES
- Applicant Address: FR Paris
- Assignee: COMMISSARIAT A L'ÉNERGIE ATOMIQUE ET AUX ÉNERGIES ALTERNATIVES
- Current Assignee: COMMISSARIAT A L'ÉNERGIE ATOMIQUE ET AUX ÉNERGIES ALTERNATIVES
- Current Assignee Address: FR Paris
- Agency: BakerHostetler
- Priority: FR1562918 20151221
- International Application: PCT/EP2016/080004 WO 20161207
- International Announcement: WO2017/108399 WO 20170629
- Main IPC: G06N3/04
- IPC: G06N3/04 ; G06N3/063

Abstract:
A neuron circuit is capable of producing a weighted sum of digitized input signals and applying an activation function to the weighted sum so as to produce a digitized activation signal as output. The circuit includes at least: one multiplier multiplying each input signal (x1 to xn) with a weight value (w1j to wnj), one accumulator accumulating the results of the multiplier so as to produce the weighted sum, and one activation unit executing the activation function. The activation unit comprises at least one shift unit and at least one saturation unit capable of approximating a non-linear activation function. The result of the approximated activation function is obtained by one or more arithmetic shifts applied to the weighted sum.
Public/Granted literature
- US20180373977A1 OPTIMIZED NEURON CIRCUIT, AND ARCHITECTURE AND METHOD FOR EXECUTING NEURAL NETWORKS Public/Granted day:2018-12-27
Information query