Added to that, in the video @TECH posted today Simon Thorpe disparaged rate coding, which GraiMatter proposes to use.Hi Dhm,
We have been tracking GraiMatter for a few years (counting the time at the other place). I pointed to GraiMatter and Hailo as potential competitors
Hi Pepsin,
We have looked at GraiMatter a couple of times in the last couple of years.
In another place, I mentioned that I thought GraiMatter and Hailo were worth keeping an eye on:
31/08/21 20:58 Post #: 55728384
If I were worried about competitors, I'd keep an eye on Hailo and GrAi Matter both of which have silicon NNs, GrAi Matter claiming a SNN.
GrAiMatter: WO2020025680A1 DATA PROCESSING MODULE, DATA PROCESSING SYSTEM AND DATA PROCESSING METHOD
https://worldwide.espacenet.com/patent/search/family/063254640/publication/WO2020025680A1?q=wo2020025680
The GraiMatter NPU:-
While GrAiMatter claim to be avoiding the von Neumann bottleneck, I don't think the apple has fallen that far from the tree.
#########################################################################
GraiMatter uses time multiplexing:
WO2020025680A1 DATA PROCESSING MODULE, DATA PROCESSING SYSTEM AND DATA PROCESSING METHOD
The paper you cited points out a problem with time multiplexing:
ShieldSquare Captcha
iopscience.iop.org![]()
page 88
On-chip learning requires precious memory and routing resources [21], which hinders scalability. On digital technologies, this problem can be sidestepped by time-multiplexing a dedicated local plasticity processor [6, 12]. The time multiplexing approach however suffers from the same caveats as a Von Neumann computer due to the separation between the SNN and the associated plasticity processor. Other promising alternatives for true local plasticity are emerging devices (Sec. 1) and related architectures (Sec. 2.1), which allow storage and computation for plasticity to occur at the same place.
Opinions and research expressed are my own as an unlicensed individual. External links are not endorsed. Do your own research or consult a licensed financial advisor.
Spiking Neural Networks: Research Projects Or Commercial Products?
Opinions differ widely, but in this space that isn’t unusual.
MAY 18TH, 2020 - BY: BRYON MOYER
![]()
Spiking Neural Networks: Research Projects or Commercial Products?
Opinions differ widely, but in this space that isn't unusual.semiengineering.com
.