BRN Discussion Ongoing

[Diogenese]

The claims set out the specifics of the invention. In this case the invention is directed to machine learning, and it does this by adding NPUs to the final layer of a NN. The final layer is where the learning takes place. The "secondary training data set" spikes could be the activation event spikes from the sensor (camera/microphone/...), the primary training data set having been provided by the model library data used in initial configuration.

Supplementary spiking neurons (NPUs) are added to the final layer (Fig 10, 1002, 1004) where Akida does its learning, presumably to incorporate newly learned features. The ALUs of Figure 1 would be involved in the step of "performing a learning function ... by performing a synaptic weight value variation ...", bearing in mind that this is for multi-bit weights and activations.


View attachment 39178

Aha! Penny's dropped. Remember when 8-bit weights were announced?

This change may be to accommodate 8-bit weights/activations.

The ALUs may be more efficient at handling the multi-bit "spikes" than the original Akida configuration.

I found this Sanskrit engraving on Eric von Dunnycan's tomb:

https://doc.brainchipinc.com/_modules/akida_models/imagenet/model_mobilenet.html
...
weight_quantization (int, optional): sets all weights in the model to have a particular quantization bitwidth except for the weights in the first layer.
Defaults to 0.
* '0' implements floating point 32-bit weights.
* '2' through '8' implements n-bit weights where n is from 2-8 bits.
activ_quantization (int, optional): sets all activations in the model to have a particular activation quantization bitwidth.
Defaults to 0.
...
input_scaling (tuple, optional): scale factor and offset to apply to
inputs. Defaults to (128, -1). Note that following Akida convention, the scale factor is an integer used as a divider.
...
© Copyright 2022, BrainChip Holdings Ltd. All Rights Reserved.

If I recall correctly, it is only the weights that are 8-bit, and only for the purpose of compatibility with 3rd party model libraries.

If there are 8-bit weights and 4-bit activations, an 8*4 matrix would be used.

 

[Gemmax]

Aha! Penny's dropped. Remember when 8-bit weights were announced?

This change may be to accommodate 8-bit weights/activations.

The ALUs may be more efficient at handling the multi-bit "spikes" than the original Akida configuration.

I found this Sanskrit engraving on Eric von Dunnycan's tomb:

https://doc.brainchipinc.com/_modules/akida_models/imagenet/model_mobilenet.html
...
weight_quantization (int, optional): sets all weights in the model to have a particular quantization bitwidth except for the weights in the first layer.
Defaults to 0.
* '0' implements floating point 32-bit weights.
* '2' through '8' implements n-bit weights where n is from 2-8 bits.
activ_quantization (int, optional): sets all activations in the model to have a particular activation quantization bitwidth.
Defaults to 0.
...
input_scaling (tuple, optional): scale factor and offset to apply to
inputs. Defaults to (128, -1). Note that following Akida convention, the scale factor is an integer used as a divider.
...
© Copyright 2022, BrainChip Holdings Ltd. All Rights Reserved.

If I recall correctly, it is only the weights that are 8-bit, and only for the purpose of compatibility with 3rd party model libraries.

If there are 8-bit weights and 4-bit activations, an 8*4 matrix would be used.

I was thinking the same thing Dodgy. Thank you for your contributions!

 

[TECH]

Great watch ... Thanks MrNick

Hey Guys,

Did you bother to check out the cars number plate ?

SEQ (Sequence) 1010 (binary) E (Edge)

Brainchips in there somewhere

 

[Fullmoonfever]

Aha! Penny's dropped. Remember when 8-bit weights were announced?

This change may be to accommodate 8-bit weights/activations.

The ALUs may be more efficient at handling the multi-bit "spikes" than the original Akida configuration.

I found this Sanskrit engraving on Eric von Dunnycan's tomb:

https://doc.brainchipinc.com/_modules/akida_models/imagenet/model_mobilenet.html
...
weight_quantization (int, optional): sets all weights in the model to have a particular quantization bitwidth except for the weights in the first layer.
Defaults to 0.
* '0' implements floating point 32-bit weights.
* '2' through '8' implements n-bit weights where n is from 2-8 bits.
activ_quantization (int, optional): sets all activations in the model to have a particular activation quantization bitwidth.
Defaults to 0.
...
input_scaling (tuple, optional): scale factor and offset to apply to
inputs. Defaults to (128, -1). Note that following Akida convention, the scale factor is an integer used as a divider.
...
© Copyright 2022, BrainChip Holdings Ltd. All Rights Reserved.

If I recall correctly, it is only the weights that are 8-bit, and only for the purpose of compatibility with 3rd party model libraries.

If there are 8-bit weights and 4-bit activations, an 8*4 matrix would be used.

 

[buena suerte :-)]

Hey Guys,

Did you bother to check out the cars number plate ?

SEQ (Sequence) 1010 (binary) E (Edge)

Brainchips in there somewhere

Nice work TECH ...

 

[thelittleshort]

Interest - familiar phrasing and poster

BRN Discussion Ongoing

... one day ... one day ...

thestockexchange.com.au

Dr. Rajan Bedi on LinkedIn: Intelligent On-Board Processing and Re-Configurable, RF…

Intelligent On-Board Processing and Re-Configurable, RF Transponders Satellite communication is becoming more intelligent and autonomous, with a single…

www.linkedin.com

 

[IloveLamp]

iCatch Technology Partners with Prophesee to Launch iEVCam

The iEVCam is a smart camera reference system designed for developers to deploy innovative event-based edge AI solutions.

bit.ly

 

[Learning]

It's just fundamental essential to have a low-power processor for "the future of mobility & Environment".

Mercedes-Benz AG on LinkedIn: #visioneqxx #mobility #electric #mercedesbenz #innovation #electric | 40 comments

The #VISIONEQXX is our commitment to a more sustainable #mobility of the future, following the ‘Design for Environment’ approach by achieving more from less on… | 40 comments on LinkedIn

www.linkedin.com

Learning 🏖

 

[TopCat]

MilliporeSigma is looking for someone. They seem quite big with 100k+ followers on LinkedIn and 6100 employees

Director of Neuromorphic Device Engineering
https://www.simplyhired.com/job/dpfS8iA_UtXJzGPNbB2LbdrPZKwxtuWsB7KbOsUtmYJi8E-8kVAL8Q

 

[GStocks123]

Cool

https://docs.edgeimpulse.com/experts/featured-machine-learning-projects/brainchip-akida-traffic-monitoring

 

[IloveLamp]

Samsung Electronics Unveils Foundry Vision in the AI Era at Samsung Foundry Forum 2023 | Samsung Semiconductor Global

Samsung Electronics Co., Ltd., a world leader in advanced semiconductor technology, today announced its latest foundry technology innovations and business strategy at the 7th annual Samsung Foundry Forum (SFF) 2023.

semiconductor.samsung.com

 

[Jasonk]

IPR Silicone IP Ltd, has anyone come across useful information? I was curious if IPR resided in a tech precinct.
After working at ARM Mauro Diamant started IPro Silicon while simultaneously working for Tiempo, Signature IP, and SiFive all of which are listed as IP vendor partners.

I feel I am missing something....

Update: I did come across an earnings guide for $5 million. Currency was not listed.

​

 

[Fullmoonfever]

I know there has been some, shall we say conjecture recently on Nviso but this popped up in a Google search & dated week or so ago.

Pushing the neuromorphic mobile EVK.

Works for me if can get some traction

Mobile Phones - NVISO

HUMAN BEHAVIOUR AI MOBILE PHONES NVISO’s Human Behaviour AI SDK allows application developers to build innovative solutions to transform our lives using AI on mobile phones. Understand people and their behavior to make autonomous devices safe, secure, and personalized for humans. Download Trial...

nviso.ai

HUMAN BEHAVIOUR AI​

MOBILE PHONES​

NVISO’s Human Behaviour AI SDK allows application developers to build innovative solutions to transform our lives using AI on mobile phones. Understand people and their behavior to make autonomous devices safe, secure, and personalized for humans.

DOWNLOAD TRIAL EVK

AI-ENABLED​

HUMAN MACHINE INTERFACES​

NVISO’s Mobile SDK provides a robust real-time human behaviour AI API, NVISO Neuro Models™ interoperable and optimised for neuromorphic computing, the ability for flexible sensor integration and placement while delivering faster development cycles and time-to-value for software developers and integrators. It enables solutions that can sense, comprehend, and act upon human behavior including emotion recognition, gaze detection, distraction detection, drowsiness detection, gesture recognition, 3d face tracking, face analysis, facial recognition, object detection, and human pose estimation. Designed for real-world environments using edge computing it uniquely targets deep learning for embedded systems,

NVISO delivers real-time perception and observation of people and objects in contextual situations combined with the reasoning and semantics of human behavior based on trusted scientific research. The NVISO Mobile SDK is supported through a long term maintenance agreement for multi-party implementation of tools for AI systems development and can be used with large-scale neuromorphic computing systems. When used with neuromorphic chips, the NVISO Mobile SDK can be used to build gaze detection systems, distraction and drowsiness detection systems, facial emotion recognition software, and a range of other applications of neuromorphic computing where understanding human behaviour in real-time is mission critical.

 

[Diogenese]

I know there has been some, shall we say conjecture recently on Nviso but this popped up in a Google search & dated week or so ago.

Pushing the neuromorphic mobile EVK.

Works for me if can get some teaction

Mobile Phones - NVISO

HUMAN BEHAVIOUR AI MOBILE PHONES NVISO’s Human Behaviour AI SDK allows application developers to build innovative solutions to transform our lives using AI on mobile phones. Understand people and their behavior to make autonomous devices safe, secure, and personalized for humans. Download Trial...

nviso.ai

HUMAN BEHAVIOUR AI​

MOBILE PHONES​

NVISO’s Human Behaviour AI SDK allows application developers to build innovative solutions to transform our lives using AI on mobile phones. Understand people and their behavior to make autonomous devices safe, secure, and personalized for humans.

DOWNLOAD TRIAL EVK

AI-ENABLED​

HUMAN MACHINE INTERFACES​

NVISO’s Mobile SDK provides a robust real-time human behaviour AI API, NVISO Neuro Models™ interoperable and optimised for neuromorphic computing, the ability for flexible sensor integration and placement while delivering faster development cycles and time-to-value for software developers and integrators. It enables solutions that can sense, comprehend, and act upon human behavior including emotion recognition, gaze detection, distraction detection, drowsiness detection, gesture recognition, 3d face tracking, face analysis, facial recognition, object detection, and human pose estimation. Designed for real-world environments using edge computing it uniquely targets deep learning for embedded systems,

NVISO delivers real-time perception and observation of people and objects in contextual situations combined with the reasoning and semantics of human behavior based on trusted scientific research. The NVISO Mobile SDK is supported through a long term maintenance agreement for multi-party implementation of tools for AI systems development and can be used with large-scale neuromorphic computing systems. When used with neuromorphic chips, the NVISO Mobile SDK can be used to build gaze detection systems, distraction and drowsiness detection systems, facial emotion recognition software, and a range of other applications of neuromorphic computing where understanding human behaviour in real-time is mission critical.

View attachment 39212

Hi Fmf,

This bit is especially interesting:

Mobile Phones - NVISO

HUMAN BEHAVIOUR AI MOBILE PHONES NVISO’s Human Behaviour AI SDK allows application developers to build innovative solutions to transform our lives using AI on mobile phones. Understand people and their behavior to make autonomous devices safe, secure, and personalized for humans. Download Trial...

nviso.ai

MICROCONTROLLER UNIT (MCU)

AI functionality is implemented in low-cost MCUs via inference engines specifically targeting MCU embedding design requirements which are configured for low-power operations for continuous monitoring to discover trigger events in a sound, image, or vibration and more. In addition, the availability of AI-dedicated co-processors is allowing MCU suppliers to accelerate the deployment of machine learning functions.

 

[Fullmoonfever]

Hi Fmf,

This bit is especially interesting:

Mobile Phones - NVISO

HUMAN BEHAVIOUR AI MOBILE PHONES NVISO’s Human Behaviour AI SDK allows application developers to build innovative solutions to transform our lives using AI on mobile phones. Understand people and their behavior to make autonomous devices safe, secure, and personalized for humans. Download Trial...

nviso.ai

MICROCONTROLLER UNIT (MCU)

AI functionality is implemented in low-cost MCUs via inference engines specifically targeting MCU embedding design requirements which are configured for low-power operations for continuous monitoring to discover trigger events in a sound, image, or vibration and more. In addition, the availability of AI-dedicated co-processors is allowing MCU suppliers to accelerate the deployment of machine learning functions.

Hi D

You thinking a potential tie in with someone in particular

 

[Diogenese]

Hi D

You thinking a potential tie in with someone in particular

Do we know where nViso gets their MCUs?

Akida 1500 could work with MCUs. All that is needed is sufficient processing power and memory to enable configuration of Akida and to handle the output from Akida.

We know Renesas has a licence for 2 nodes (8 NPUs), but I don't know how many nodes are required for nViso's functions like face recognition.

nViso is a BrainChip partner, so it may not need a licence:

BrainChip and NVISO are targeting battery-powered applications in robotics and mobility devices addressing the need for high levels of AI performance in ultra low power environments. Implementing NVISO’s AI solutions with BrainChip’s Akida drive next generation solutions.



This is another signpost to Akida:

NVISO NEURO MODELS™​

ULTRA-EFFICIENT DEEP LEARNING AT THE EDGE​

NVISO Neuro Models™ are purpose built for a new class of ultra-efficient machine learning processors designed for ultra-low power edge devices. Supporting a wide range of heterogenous computing platforms ranging from CPU, GPU, DSP, NPU, and neuromorphic computing they reduce the high barriers-to-entry into the AI space through cost-effective standardized AI Apps which work optimally at the extreme edge (low power, on-device, without requiring an internet connection). NVISO uses low and mixed precision activations and weights data types (1 to 8-bit) combined with state-of-the-art unstructured sparsity to reduce memory bandwidth and power consumption. Proprietary compact network architectures can be fully sequential suitable for ultra-low power mixed signal inference engines and fully interoperable with both GPUs and neuromorphic processors.

 

[Fullmoonfever]

Recent Indian article on MBRDI in India.

Neuromorphic and ChatGPT anyone?

Couple of excerpts.

How Mercedes-Benz is driving a culture of innovation at its research and development centre By embracing innovation, the company is empowering its workforce to adopt a mindset that encourages thinking beyond conventional boundaries

By Radhika Sharma | HRKatha -June 23, 2023

Read more at: https://www.hrkatha.com/features/how-mercedes-benz-is-driving-a-culture-of-innovation/

Mercedes-Benz Research and Development India (MBRDI) recognises the pivotal role of innovation in promoting a healthy work culture. “One aspect that sets us apart is our work culture and emphasis on employee well-being. As the largest research and development centre of Mercedes-Benz outside of Germany, with a growing workforce, we have established ourselves as a powerhouse for delivering cutting-edge technology and innovation. Our primary focus is to drive the future of sustainable mobility, ensuring high quality, safety and innovation. Our commitment to innovation extends beyond technology and is reflected in our progressive work culture,” points out Mahesh Medhekar, vice president-human resources, Mercedes-Benz Research and Development India.


Innovation learning: The company focuses on innovation learning for its employees. Recently, it hosted a seminar on biomimicry — a field that combines generative AI and neuromorphic computing to emulate the human brain’s behaviour. This cutting-edge science allows MBRDI to enhance the capabilities of its vehicles, enabling them to understand and respond to the driver’s behaviour, ensuring safety and comfort. “Innovation learning encompasses various aspects, including deep learning, software patterns and techniques such as design thinking, all aimed at fostering innovation within our organisation. These practices serve as key pillars in our pursuit of continuous innovation,” enunciates Medhekar.

“Innovation learning encompasses various aspects, including deep learning, software patterns and techniques such as design thinking, all aimed at fostering innovation within our organisation. These practices serve as key pillars in our pursuit of continuous innovation.” Mahesh Medhekar, vice president-human resources, Mercedes-Benz Research and Development India

 

[Fullmoonfever]

Do we know where nViso gets their MCUs?

Akida 1500 could work with MCUs. All that is needed is sufficient processing power and memory to enable configuration od Akida and to handle the output from Akida.

We know Renesas has a licence for 2 nodes (8 NPUs), but I don't know how many nodes are required for nViso's functions like face recognition.

nViso is a BrainChip partner, so it may not need a licence:


View attachment 39219

BrainChip and NVISO are targeting battery-powered applications in robotics and mobility devices addressing the need for high levels of AI performance in ultra low power environments. Implementing NVISO’s AI solutions with BrainChip’s Akida drive next generation solutions.



This is another signpost to Akida:

NVISO NEURO MODELS™​

ULTRA-EFFICIENT DEEP LEARNING AT THE EDGE​

NVISO Neuro Models™ are purpose built for a new class of ultra-efficient machine learning processors designed for ultra-low power edge devices. Supporting a wide range of heterogenous computing platforms ranging from CPU, GPU, DSP, NPU, and neuromorphic computing they reduce the high barriers-to-entry into the AI space through cost-effective standardized AI Apps which work optimally at the extreme edge (low power, on-device, without requiring an internet connection). NVISO uses low and mixed precision activations and weights data types (1 to 8-bit) combined with state-of-the-art unstructured sparsity to reduce memory bandwidth and power consumption. Proprietary compact network architectures can be fully sequential suitable for ultra-low power mixed signal inference engines and fully interoperable with both GPUs and neuromorphic processors.

Wondered if you were fishing in Renesas' pond

Not inconceivable....they taped out for an unnamed 3rs party

 

[Diogenese]

Wondered if you were fishing in Renesas' pond

Not inconceivable....they taped out for an unnamed 3rs party

But just to muddy the waters, Akida is qualified across ARMs range of processors, not to mention Intel, MegaChips, ...

 

[Fullmoonfever]

Geez....we might need a bigger bandwagon

On May 2023, EDI started the implementation of a new Horizon Europe project (EcoMobility) with the goal to develop a connected transportation ecosystem – Elektronikas un datorzinātņu institūts

On May 2023, EDI started the implementation of a new Horizon Europe project (EcoMobility) with the goal to develop a connected transportation ecosystem​

ELEKTRONIKAS UN DATORZINĀTŅU INSTITŪTS > NEWS > ON MAY 2023, EDI STARTED THE IMPLEMENTATION OF A NEW HORIZON EUROPE PROJECT (ECOMOBILITY) WITH THE GOAL TO DEVELOP A CONNECTED TRANSPORTATION ECOSYSTEM
30. June, 2023

“Improved automotive vision systems are a crucial stepping stone towards safe and sustainable transportation, hence developing reliable and energy efficient solutions is in the center of attention. To address this challenge, the field of neuromorphic computing, a brain-inspired branch of AI, holds great promise. Specifically, spiking neural networks (SNNs), which mimic the structure and functionality of the nervous system, offer notable advantages such as speed and energy efficiency. Until now, the lack of efficient hardware accelerators has hindered the widespread adoption of SNNs. Nevertheless, this situation is rapidly changing, and EDI, with its expertise in reconfigurable logic devices, is well-positioned to contribute to further advancements in this field of research,”

on new frontiers in neuromorphic machine perception reflects Edgars Lielāmurs, the research assistant at EDI.