BRN Discussion Ongoing

Diogenese

Top 20
We have established BrainChip Systems India to support the requirement for robust system software and firmware. As Akida is implemented commercially, it is important that our software is mature and provides a positive user experience.”
We have software for Akida Development Environment (ADE) and MetaTF.

We also need some software/firmware to enable our ARM Cortex microprocessor to be able to configure the NPUs and manage the weights for the NPUs.

There is also software used in the CNN2SNN conversion.

https://doc.brainchipinc.com/user_guide/cnn2snn.html

By careful attention to specifics in the architecture and training of the CNN, an overly complex conversion step from CNN to SNN can be avoided. The CNN2SNN toolkit comprises a set of functions designed for the popular Tensorflow Keras framework, making it easy to train a SNN-compatible network.

1647169742416.png




Typical training scenario

The first step in the conversion workflow is to train a standard Keras model. This trained model is the starting point for the quantization stage. Once it is established that the overall model configuration prior to quantization yields a satisfactory performance on the task, we can proceed with quantization.

The CNN2SNN toolkit offers a turnkey solution to quantize a model: the quantize function. It replaces the neural Keras layers (Conv2D, SeparableConv2D and Dense) and the ReLU layers with custom CNN2SNN layers, which are quantization-aware derived versions of the base Keras layer types. The obtained quantized model is still a Keras model with a mix of CNN2SNN quantized layers (QuantizedReLU, QuantizedDense, etc.) and standard Keras layers (BatchNormalization, MaxPool2D, etc.).

Direct quantization of a standard Keras model (also called post-training quantization) generally introduces a drop in performance. This drop is usually small for 8-bit or even 4-bit quantization of simple models, but it can be very significant for low quantization bitwidth and complex models.

If the quantized model offers acceptable performance, it can be directly converted into an Akida model, ready to be loaded on the Akida NSoC (see the convert function).

However, if the performance drop is too high, a quantization-aware training is required to recover the performance prior to quantization. Since the quantized model is a Keras model, it can then be trained using the standard Keras API.

Note that quantizing directly to the target bitwidth is not mandatory: it is possible to proceed with quantization in a serie of smaller steps. For example, it may be beneficial to keep float weights and only quantize activations, retrain, and then, quantize weights.


When Akida is processing input data, the software is dormant.

But when we get to Akida 3000's neural cortex, who knows?
 
  • Like
  • Fire
Reactions: 25 users
We have software for Akida Development Environment (ADE) and MetaTF.

We also need some software/firmware to enable our ARM Cortex microprocessor to be able to configure the NPUs and manage the weights for the NPUs.

There is also software used in the CNN2SNN conversion.

https://doc.brainchipinc.com/user_guide/cnn2snn.html

By careful attention to specifics in the architecture and training of the CNN, an overly complex conversion step from CNN to SNN can be avoided. The CNN2SNN toolkit comprises a set of functions designed for the popular Tensorflow Keras framework, making it easy to train a SNN-compatible network.

View attachment 2509



Typical training scenario

The first step in the conversion workflow is to train a standard Keras model. This trained model is the starting point for the quantization stage. Once it is established that the overall model configuration prior to quantization yields a satisfactory performance on the task, we can proceed with quantization.

The CNN2SNN toolkit offers a turnkey solution to quantize a model: the quantize function. It replaces the neural Keras layers (Conv2D, SeparableConv2D and Dense) and the ReLU layers with custom CNN2SNN layers, which are quantization-aware derived versions of the base Keras layer types. The obtained quantized model is still a Keras model with a mix of CNN2SNN quantized layers (QuantizedReLU, QuantizedDense, etc.) and standard Keras layers (BatchNormalization, MaxPool2D, etc.).

Direct quantization of a standard Keras model (also called post-training quantization) generally introduces a drop in performance. This drop is usually small for 8-bit or even 4-bit quantization of simple models, but it can be very significant for low quantization bitwidth and complex models.

If the quantized model offers acceptable performance, it can be directly converted into an Akida model, ready to be loaded on the Akida NSoC (see the convert function).

However, if the performance drop is too high, a quantization-aware training is required to recover the performance prior to quantization. Since the quantized model is a Keras model, it can then be trained using the standard Keras API.

Note that quantizing directly to the target bitwidth is not mandatory: it is possible to proceed with quantization in a serie of smaller steps. For example, it may be beneficial to keep float weights and only quantize activations, retrain, and then, quantize weights.


When Akida is processing input data, the software is dormant.

But when we get to Akida 3000's neural cortex, who knows?
Precisely that’s why I am being very nice to AKD1000, 500, 1500 & 2000 just in case. I don’t think I could sleep with the lights out if I had made that remark about “who knows what a neuromorphic chip is anyway.” FF
 
  • Like
  • Haha
Reactions: 11 users

VictorG

Member
Good morning chippers, wishing each of you a great week.

Yesterday I was at a family function together with many family friends, I intruding many of them to BRN. Collectively we have purchased just shy of 2 million BRN shares over the past 7 months. The mood was upbeat and most were inclined to add to their holdings at these levels.

I asked a mate why he so bullish on BRN, he said the only price he sees is his target price in 2 yrs, everything else is just noise.

GLTA

VG
 
  • Like
  • Fire
  • Love
Reactions: 32 users

Tezza

Regular
Good morning chippers, wishing each of you a great week.

Yesterday I was at a family function together with many family friends, I intruding many of them to BRN. Collectively we have purchased just shy of 2 million BRN shares over the past 7 months. The mood was upbeat and most were inclined to add to their holdings at these levels.

I asked a mate why he so bullish on BRN, he said the only price he sees is his target price in 2 yrs, everything else is just noise.

GLTA

VG
I am very similar to your friend, i have a 2 yr target price of $8. Might be optimistic with the current going ins in the world so I may need to stretch to 3 yrs.
 
  • Like
Reactions: 12 users

chapman89

Founding Member
I am very similar to your friend, i have a 2 yr target price of $8. Might be optimistic with the current going ins in the world so I may need to stretch to 3 yrs.
In 2 years time we will have AKD2000 out of production, consistent flow of revenue, multiple contracts in various industries, most likely will have many US retail and institutional investors buying up shares, I think $8 is conservative for 2 years time.
 
  • Like
  • Fire
Reactions: 61 users
D

Deleted member 118

Guest

Attachments

  • 08B79A71-D2FB-459E-AD2E-47035B55A8EB.png
    08B79A71-D2FB-459E-AD2E-47035B55A8EB.png
    546.4 KB · Views: 153
Last edited by a moderator:
  • Like
  • Fire
  • Love
Reactions: 40 users
Interesting, the invasion may spark a China invasion of Taiwan.

ABC News: Taipei residents warn against the idea that 'Taiwan is next' after Russia's invasion of Ukraine.

Good news though:

"The US has the Taiwan Relations Act, which states, "The United States will make available to Taiwan such defense articles and defense services in such quantity as may be necessary to enable Taiwan to maintain a sufficient self-defense capability."

Analysts describe this as "strategic ambiguity" when it comes to what kind of support the US would provide in the event China invaded Taiwan.

But in October last year, when asked if the US would come to Taiwan's defence if China attacked, President Joe Biden responded, "Yes, we have a commitment to do that."
 
  • Like
Reactions: 9 users

Mccabe84

Regular
I plan on buying more over the next couple of months because Sean said yes there will be revenue this year, so surely the share price would be higher than it is now with revenue coming in
 
  • Like
  • Thinking
Reactions: 14 users
IMPORTANT:

Make sure you take 5 minutes to read the @uiux

BrainChip + Quantum Ventura​

thread. It will be worth the effort.

My opinion only DYOR
FF

AKIDA BALLISTA
 
  • Like
  • Fire
Reactions: 28 users

Motty

Member
Yes.
Has been discussed.

They claim their aim is to compete with Nvidia. They have three boards ranging in power consumption from 75 watts to 300 watts. They do deep training neural networks and are chasing software development at the moment to support their systems.

They do not do SNN or low power or Ai/AGi.

They call their current product Greyscale.

My opinion only DYOR
FF

AKIDA BALLISTA

PS: More likely to partner with Brainchip than compete just like Nvidia.
Thanks for the reply FF.
 
  • Like
Reactions: 2 users
D

Deleted member 118

Guest

Attachments

  • 31DDDE24-88F4-4052-A04F-313AF92F85CE.png
    31DDDE24-88F4-4052-A04F-313AF92F85CE.png
    7.1 MB · Views: 105
Last edited by a moderator:
  • Like
Reactions: 3 users
The other day Ford’s brain machine interface was discussed with the discovery of its patent for detecting driver alertness and reaction from Brain activity and or muscle response now over on the Mercedes thread Alfie Posted the following link for Mercedes:


I said it the other day the only way this can happen efficiently is with a processor that can recognise spike patterns. We know this is how Neuralink is doing this for Mr. Musk and as I said the crossovers have led many to wonder if Neuralink might be using AKIDA technology.

Is it not just too much of a coincidence that both Ford and Mercedes are doing similar things at the same time as they are working with Brainchip.

As Dio said and all this for just $1.05.

True value should be $1.50 according to Pitt Street Research and since then we have had at least the following:
1. Almost a doubling of staff and larger offices and more staff to come;
2. The release of four commercial ready to buy COTS;
3. The new CEO Sean Hehir
4. The new Head of Marketing Jerome Nadel
5. The confirmation that all 15 EAPs look likely to convert;
6. The licence with MegaChips
7. The very successful capital call on LDA Capital;
8. A great looking annual report showing $49 million in the bank;
9. A 4C showing revenue increasing;
10. Validation unsolicited from Mercedes Benz;
11. Confirmation that Mercedes Benz, Valeo, NASA & Vorago have moved from being simple EAPs to being Early Adopters;
12. Confirmation that Renesas will be releasing MCUs this year with AKIDA technology on board;
13. Go to market readiness of MegaChips with trained staff out there selling AKIDA edge technology;
14. A new Chair of the Board Mr. Viana completing the Brainchip ARM management and strategic plan;
15. The appointment of a highly qualified Australian director;
16. Confirmation that the AKD2000 design is complete and on schedule;
17. The reveal that Brainchip has been awarded research and development grants by the Western Australian and Federal Governments and has had an endorsement by the WA Chief Scientist; and
18. The research grant from the US Airforce for the AKIDA radar project.

Clearly at an average value of 2 cents a point x18 points BRNASX should now be sitting at $1.86 at least.

$1.05 cents is irrational.

My opinion only DYOR
FF

AKIDA BALLISTA
It is now after 10am on Monday and trading has commenced. True to my word I did not debate the comments made by any other poster regarding their views as to whether $1.05 was a rational valuation of BRNASX or not and based on the responses in this straw poll I can conclude:

1. My fair value assessment of $1.88 is incorrect and in the opinion of all those who replied far too low;

2. Minimum fair value is $2.00 at least;

3. Present valuations on the ASX are irrational;

4. I am a downramper sleeper sent by HC to play the long game.

My opinion only DYOR
FF

AKIDA BALLISTA
 
  • Like
  • Haha
  • Love
Reactions: 39 users
D

Deleted member 118

Guest

Attachments

  • BD2CDFBD-3EF2-4ADF-94E4-623F1D1CD143.png
    BD2CDFBD-3EF2-4ADF-94E4-623F1D1CD143.png
    1.1 MB · Views: 107
Last edited by a moderator:
  • Like
  • Fire
Reactions: 25 users

Labsy

Regular
Brainchip shows up in Yahoo and Safari searches, first page.
All someone here with technical know how needs to do is update Wikipedia to include neuromorphic processors, ie Brainchip being first in market, and problem solved.
I googled one shot learning and a complex Wikipedia explanation i suspect on algorithms came up, top of the page ;)
 
  • Like
  • Fire
Reactions: 13 users
D

Deleted member 118

Guest

Attachments

  • D21F710F-7EF3-45A7-9965-528BD0A8F746.png
    D21F710F-7EF3-45A7-9965-528BD0A8F746.png
    505.3 KB · Views: 104
Last edited by a moderator:
  • Like
  • Fire
Reactions: 11 users

M_C

Founding Member
  • Like
  • Fire
  • Wow
Reactions: 53 users

yogi

Regular
Hi everyone hope you all had nice weekend.
Was just going through the Sale side and something weird happening today. Once again there are sale of 1 or 2 or 3 shares in one transaction. I went through so see how many and there has been plenty today sale of under 500$ what does that mean is someone trying to harm SP.
If i want to buy less then 500$ worth or shares i can't.
Seems plenty people are playing some sort of game or what.
cheers
 
  • Like
Reactions: 11 users

davidfitz

Regular
Hi everyone hope you all had nice weekend.
Was just going through the Sale side and something weird happening today. Once again there are sale of 1 or 2 or 3 shares in one transaction. I went through so see how many and there has been plenty today sale of under 500$ what does that mean is someone trying to harm SP.
If i want to buy less then 500$ worth or shares i can't.
Seems plenty people are playing some sort of game or what.
cheers
As long as it stays over $1 I suppose I am happy although it should be more on what we all know. The games have been going on for a while now so most of us have just begrudgingly accepted it.

Whilst it would be easy to blame the events overseas the NDA theme is starting to sound a bit like a broken record. And yes I do understand the need for them but it is growing increasingly frustrating.

Despite the continued optimism by most of us LTHs if it falls under this 'psychological' figure I think in general we lose some credibility with the overall market.

I personally am coming up to my 7th anniversary with my first purchase and I had dreamt of more by this stage. Maybe I was a bit unrealistic with my expectations but some of us can't hold onto their investment forever. It was nice to be a millionaire briefly at the beginning of the year so at least I have that to be thankful for. Who knows maybe it will happen again in the near future.
 
  • Like
  • Fire
  • Love
Reactions: 35 users

VictorG

Member
As long as it stays over $1 I suppose I am happy although it should be more on what we all know. The games have been going on for a while now so most of us have just begrudgingly accepted it.

Whilst it would be easy to blame the events overseas the NDA theme is starting to sound a bit like a broken record. And yes I do understand the need for them but it is growing increasingly frustrating.

Despite the continued optimism by most of us LTHs if it falls under this 'psychological' figure I think in general we lose some credibility with the overall market.

I personally am coming up to my 7th anniversary with my first purchase and I had dreamt of more by this stage. Maybe I was a bit unrealistic with my expectations but some of us can't hold onto their investment forever. It was nice to be a millionaire briefly at the beginning of the year so at least I have that to be thankful for. Who knows maybe it will happen again in the near future.
I understand the frustration and yes it can weigh us down. I resolve in the fact that without any doubt, nothing can stop BRN's eventual rapid price rise, be it in 1 month or 1 yr. I too will go back to millionaire status while by my calculations you will go back to multi-millionaire status.
 
  • Like
  • Fire
Reactions: 17 users

Esq.111

Fascinatingly Intuitive.
As long as it stays over $1 I suppose I am happy although it should be more on what we all know. The games have been going on for a while now so most of us have just begrudgingly accepted it.

Whilst it would be easy to blame the events overseas the NDA theme is starting to sound a bit like a broken record. And yes I do understand the need for them but it is growing increasingly frustrating.

Despite the continued optimism by most of us LTHs if it falls under this 'psychological' figure I think in general we lose some credibility with the overall market.

I personally am coming up to my 7th anniversary with my first purchase and I had dreamt of more by this stage. Maybe I was a bit unrealistic with my expectations but some of us can't hold onto their investment forever. It was nice to be a millionaire briefly at the beginning of the year so at least I have that to be thankful for. Who knows maybe it will happen again in the near future.
Afternoon Chippers,

Jolly good dot joining all.

Just a reminder that the first of the PCI Cards should be received by those that ordered one , Thursday / Friday this week.

It will have been 8 weeks since announcement ......... to........ DELIVERY.

Hopefully start to see some new applications of our tech from those among us here and those further afield.

Particularly like to see a photo of one nailed to the wall in some ones billiard room.

Regards,
Esq.
 
  • Like
  • Love
  • Fire
Reactions: 31 users
Top Bottom