We added CNN2SNN and 4-bit weights/activations, and are adding LSTM (long short term memory) in Akida 1500.
CNN is used for image processing.
4-bit weights/activations are more accurate then 1-bit W/As, at the expense of a little speed and a little power usage increase.
LSTM is used for sequence learning networks.
BrainChip – Annual General Meeting CEO and Chairman’s Address
Sydney 26 May 2021
https://hotcopper.com.au/data/announcements/ASX/2A1300009_BRN.pdf
So who wants more accurate, fairly fast, fairly power efficient image sequence processing?
2022 BRN Discussion, page-801
-
- There are more pages in this discussion • 10,122 more messages in this thread...
You’re viewing a single post only. To view the entire thread just sign in or Join Now (FREE)
Add BRN (ASX) to my watchlist
 (20min delay)
|
|||||
|
Last
16.0¢ |
Change
-0.005(3.03%) |
Mkt cap ! $313.9M | |||
| Open | High | Low | Value | Volume |
| 17.0¢ | 17.0¢ | 16.0¢ | $765.2K | 4.687M |
Buyers (Bids)
| No. | Vol. | Price($) |
|---|---|---|
| 107 | 2575551 | 16.0¢ |
Sellers (Offers)
| Price($) | Vol. | No. |
|---|---|---|
| 16.5¢ | 873768 | 15 |
View Market Depth
| Last trade - 16.10pm 03/09/2024 (20 minute delay) ? |
| BRN (ASX) Chart |