1. 
   

   
   

   1. a digressive series of post and can be skipped.
      
      Deep learning in neural networks: An overview
      

      • Jürgen Schmidhuber

      
      
      Abstract
      In recent years, deep artificial neural networks (including recurrent ones) have won numerous contests in pattern recognition and machine learning. This historical survey compactly summarizes relevant work, much of it from the previous millennium. Shallow and Deep Learners are distinguished by the depth of their credit assignment paths, which are chains of possibly learnable, causal links between actions and effects. I review deep supervised learning (also recapitulating the history of backpropagation), unsupervised learning, reinforcement learning & evolutionary computation, and indirect search for short programs encoding deep and large networks.

       

      jfieb, 9 minutes agoEditDeleteReport

      #19Reply
   2. 

      jfiebMember

      New

      Neural Networks for Machine Learning
      Learn about artificial neural networks and how they're being used for machine learning, as applied to speech and object recognition, image segmentation, modeling language and human motion, etc. We'll emphasize both the basic algorithms and the practical tricks needed to get them to work well.
      
      
      
      About the Course
      Neural networks use learning algorithms that are inspired by our understanding of how the brain learns, but they are evaluated by how well they work for practical applications such as speech recognition, object recognition, image retrieval and the ability to recommend products that a user will like. As computers become more powerful, Neural Networks are gradually taking over from simpler Machine Learning methods. They are already at the heart of a new generation of speech recognition devices and they are beginning to outperform earlier systems for recognizing objects in images. The course will explain the new learning procedures that are responsible for these advances, including effective new proceduresr for learning multiple layers of non-linear features, and give you the skills and understanding required to apply these procedures in many other domains.
      
      This YouTube video gives examples of the kind of material that will be in the course, but the course will present this material at a much gentler rate and with more examples.
      
      
      QUIK has a hardcoded FFE, they hard coded a small part of Sensory audio algos, do NOT be surprised that since learning algos will be Ubiquitous and since Ubiquity = a future Eos Engine that QUIK is working on a hard code neural network.
      
      It is part of the adjacent possible that QUIK is exploring. FFE = Flexible Fusion engine.
      
      This one will be NNLE -neural network learning engine. It will be ubiquitous

There will be tiers of intelligence that stay on the device.

Its been a vision of Sensory for yrs.

Its is such a good fit for what QUIK has.

Tiers of intelligence that stays local is a new conclusion, 
Its a fork in the road...many will HAND it ALL into the cloud- as that is what they have on their benches.

Sensory/QUIK are a great match.
Local intelligence will maintain margin and it will start to separate the horses in the race.  Many can't go here.  Nice.

Principal, Algorithm Architect
	Employment Type :Full Time Regular Job ClassProfessional
The principal sensor algorithm architect plays a key role indeveloping advanced sensor algorithms on QuickLogic’s sensor hub for smartphone and wearable.  Strong algorithm development andsoftware implementation experience, with the ability to collaborate withmulti-disciplined engineering teams, is required for success in this position. Inthis role, he or she is responsible for developing, enhancing and maintaining advancedsensor algorithms including motion sensor calibration and fusion, contextawareness, gesture recognition and other sensor algorithms on QuickLogic’ssmart-sensor hub for smartphones and wearable. He or she is responsible for allphases of development including generating algorithm specifications based on keyperformance indicators and other market requirements,algorithmsimulation, real-time implementation using C and Assembly, performance tuningfor various use cases, optimization for memory footprint and computation load. He or she is also responsible for guiding thetesting engineer for generating the testing methods and apparatus for sensoralgorithms, sensor data collection, visualization and annotation for sensor algorithms. The engineer will support for key customer engagement toensure design wins and successful integration, qualification and deployment ofthe smart sensor products. He or she may also engage with our partners,customers, and other engineering teams for generating the technical requirementsand generating the system and algorithm specifications Skills & Knowledge Profile PhDwith 7+ years in electrical engineering or computer science is preferred. MSwith 10+ years in electrical engineering or computer science is required.   Expertlevel in algorithm development using sensor fusion, machine learning, patternrecognition, simulation and mathematical analysis. Familiar with algorithmdevelopment tools such as Matlab and Simulink. Experiencein efficient implementation of floating point and fixed point algorithms in firmware,software and hardware. Demonstratedability to successfully complete complex R&D technology projects Mustbe able to work in team environment The following experience is highly desired: Androidsystem architecture experience + the phrase neural network is just a good fit with a company that has reconfigurable silicon? This is the best stuff as far as digressive reading in the past yr or more.

Sensory decided to keep it on the device yrs ago. QUIK focused on low power algo crunching.

For the casual reader...this the the BEST thing, its a new conclusion- A MAJOR concern. Its not just Dr. Saxe talking-but Sensory Inc vision also. Just wait for EOS 2 as it will add MORE intelligence.

Would I buy a part of Sensory's business? You bet.....
QUIK for me IS more diversified 5-10% of my investment is now an investment in things Sensory is doing....


QUIK IS at the cutting edge in being able to implement deeply embedded machine learning. Sensory is worth a LOT, hope they stay independent.

New phrases that are where the EOs 2 will be...

intelligence, machine learning, neural networks, on device,...it take us into the artifical intelligence realm.

The Adjacent Possible says that QUIK will have a long term relationship with Sensory Inc.,

here is their most recent blog

GUEST BLOG – RISE OF THE MACHINES (LEARNING)

November 12, 2015

A really smart guy told me years ago that neural networks would prove to be the second best solution to many problems.  While he was right about lots of stuff, he missed that one!  Out of favor for years, neural networks have enjoyed a resurgence fueled by advances in deep machine learning techniques and the processing power to implement them.  Neural networks are now seen to be the leading solution to a host of challenges around mimicking how the brain recognizes patterns.

Google’s Monday announcement that it was releasing its TensorFlow machine learning system on an open-source basis underscores the significance of these advances, and further validates Sensory’s 22 year commitment to machine learning and neural networks.  TensorFlow is intended to be used broadly by researchers and students “wherever researchers are trying to make sense of very complex data — everything from protein folding to crunching astronomy data”.  The initial release of TensorFlow will be a version that runs on a single machine, and it will be put into effect for many computers in the months ahead, Google said.

Microsoft also had cloud-based machine learning news on Monday, announcing an upgrade to Project Oxford’s facial recognition API launched in May specifically for the Movember Foundation’s no-shave November fundraising effort: a facial hair recognition API that can recognize moustache and beard growth and assign it a rating (as well as adding a moustache “sticker” to the faces of facial hair posers).

Project Oxford’s cloud-based services are based on the same technology used in Microsoft’s Cortana personal assistant and the Skype Translator service, and also offer emotion recognition, spell check, video processing for facial and movement detection, speaker recognition and custom speech recognition services.

While Google and Microsoft have announced some impressive machine-learning capabilities in the cloud, Sensory uniquely combines voice and face for authentication and improved intent interpretation on device, complementing what the big boys are doing.

From small footprint neural networks for noise robust voice triggers and phrase-spotted commands, to large vocabulary recognition leveraging a unique neural network with deep learning that achieves acoustic models an order of magnitude smaller than the present state-of-the-art, to convolutional neural networks deployed in the biometric fusion of face and voice modalities for authentication, all on device and not requiring any cloud component, Sensory continues to be the leader in utilizing state-of-the-art machine learning technology for embedded solutions.

Not bad company to keep!

Bernard Brafman
Vice President of Business Development

Email

inShare

Categories: Biometrics, Security, voice authentication
Tags: facial recognition, google, Microsoft, neural networks, Project Oxford, TensorFlow,voice authentication

Comments Off on Guest Blog – Rise of the Machines (Learning)

here is a cc paragraph to read again…


Today, we have three active SenseMe license engagements with large OEMs. All three OEMs have high brand-name recognition and two are sector leaders. I am very pleased to announce that we have received our first purchase order for a SenseMe license from one of these OEMs and we expect to announce further details on the design once the customer releases its new product in conjunction with CES. I can't get into the detail right now, but I am excited about the number of new products using QuickLogic's sensor processing solutions that will be displayed at CES in 2016. If you plan to attend CES in January, I encourage you to make an appointment to visit us.


So there is reason to look forward to CES, and the teasers are starting already.

 

HomeForums>Main Category>Main Forum>

Misfit goes to Fossil for 250 mil

1. • Fossil acquires wearable maker Misfit for $260 million

   
   Watchmakers and wearables unite
   

   • 
     

   
   
   Share on Facebook Tweet (432) Share (207) Pin (1)
   Fossil Group has agreed to acquire Misfit, maker of wearable activity trackers for $260M in a major deal uniting traditional watchmakers with Silicon Valley talent. Fossil Group says it plans to begin incorporating Misfit's technology into products resembling traditional watches as early as next year.
   
   FOSSIL GROUP SELLS 50 MILLION WATCHES A YEAR
   
   Since it was founded in 2011, Misfit has built sleek, inexpensive activity monitors — like the Shine 2, which lasts up to six months on a coin battery — and some smarthome products like sleep monitors and smart bulbs. The company found success in partnerships with companies like Speedo and Swarovski, but the company hadn't been able to overtake industry staples like Fitbit or Nike who dominate the wearable market.
   
   "If you don’t have a brand it is hard to be legit in this space," Sonny Vu, the co-founder and CEO of Misfit told The Wall Street Journal. Vu — who has held the belief that wearables should be attractive as well as functional — will take over as president and CTO of Fossil Group, which produces watches for brands including Adidas, Emporio Armani, Michael Kors, Burberry, and DKNY. By partnering with Fossil Group, a company that sells 50 million watches a year, Misfit will gain that brand, and its technology could find a much broader audience than it has enjoyed up unto this point.
   
   "With the acquisition of Misfit, Fossil Group will be uniquely positioned to lead the convergence of style and technology and to become the fashion gateway to the high-growth wearable technology and connected device markets." Kosta Kartsotis, CEO of Fossil Group said in a statement.
   
   Apple's entry into the smartwatch market may have been a wakeup call for traditional watch companies, but Tag Heuer getting into the smartwatch game and Fossil acquiring a wearable company, the watch industry may be finally ready to embrace technology more than it ever has.

    

   jfieb, 21 minutes agoEditReport

   #1Reply
2. 

   jfiebMember

   Teardown: Misfit Shine Activity Tracker
   
   
   
   More than just fancy pedometers, a whole new class of mobile devices are flooding the market, designed to make us more aware of our own physical activity. And in this emerging market jam-packed with competitors like Nike, Garmin, and Fitbit, the Misfit Shine is easily the… prettiest. Hardly bigger than a US Quarter and barely thicker than a pencil, it can be worn anywhere on the body and looks more like a piece of jewelry than a high tech activity tracker. But can you really fit a fully functional fitness tracker in something that size? There’s only one way to find out…
   
   Don't Open 'Til Doomsday
   Getting into the Shine turns out not to be that hard. After popping open the case like you’re changing the battery, anything sharp will pop the retaining ring off of the pcb and give you access to this:
   
   
   
   Wow! That’s a lot of stuff in such a small package!
   
   There are really three parts to the Shine, everything else on the board is essentially just there to support these three. Let’s examine each part and how it helps the Shine track your fitness:
   
   Brains of the Operation
   The EFM32 Leopard Gecko microcontroller from Silicon Labs is the controller that keeps this whole operation in one piece. The Leopard Gecko is a 32-bit microcontroller based on an ARM Cortex-M3 core. It comes packed with energy saving features that make it ideal for this kind of application.
   
   
   Of particular note is the Leopard Gecko’s low-energy sensor interface which plays a big role in the Shine’s advertised 4-month battery life. The low-energy sensor interface allows the EFM32’s peripherals to communicate independently of the core (gathering accelerometer data, for instance) and allowing the core to stay in energy saving mode.
   
   
   
   The Leopard Gecko’s 256k of program storage and 48MHz processing speed are also well suited to this application since the Shine not only has to store and process sensor data using complex algorithms, but also has to communicate that processed data using the (somewhat sizable) Bluetooth low energy software stack.
   
   A Sense of Purpose
   Of course just wearing a microprocessor on your sleeve won’t tell you a lot about your fitness. To get a handle on what you’re up to, the Shine relies on the LIS3DH Accelerometer from STMicro, another ultra low-power part. This tiny LGA-16 package saves real estate on the PCB but still serves up acceleration data at speeds up to 5kHz.
   
   Communication is Key
   Collecting and processing sensor data is a neat trick, but, if you can’t ever get to that data, what good is it? Transmitting the data wirelessly is the perfect way to get it out of the device, but it takes a lot of power to send things over air. To solve this problem, Misfit is taking advantage of the brand-newish Bluetooth Low Enegry (BLE) protocol. The device behind that ability is the CC2541 Bluetooth SoC by Texas Instruments. With its programmable output power and low energy operation, the CC2541 does its part to keep the battery healthy too.
   
   All Together Now
   If I had to guess what’s going on in there, which I do, I’d say that the Leopard Gecko is spending 98% of its time in power saving mode, only waking up on boot, when syncronizing, and after detecting a double or triple tap event. The accelerometer data probably gets processed either at synchronize time or when some buffer gets filled with raw data… maybe both? Either way, I’m almost sure that your phone never receives raw accelerometer data.
   
   People who have reviewed the Shine online say that it doesn’t usually live up to its 4 month battery life promise, but it still lasts impressively long on a standard coin cell battery. Using low energy parts and keeping everything in power-save mode as long as possible has a lot to do with that.
   
   The Unexplained
   I tried to find some way of hacking the firmware on this, but I didn’t have any luck identifying a point of entry. I probed every test pin and package lead that I could reach, but I couldn’t find anything interesting on the logic analyzer. The controller, being a BGA package, wasn’t very accessible from a hardware hacking standpoint. Ultimately I gave up hope on trying to figure things out on the device side.
   
   Because the Shine is a Bluetooth Low Energy device, however, I figured it might be possible to get some information on the host side. I downloaded a utility for my smartphone that reads BLE device attributes and connected to the Shine. Unfortunately, there didn’t seem to be any way for me to access the sensor data form that end either. For anyone interested, though, here’s what I was was able to find out using the BLE utility…
   
   After connecting to the device, you have access to 4 services, which are basically categories of attributes that can be read, written, or otherwise manipulated. Generic Access, Generic Attribute, and Device Information are all standard services on the BLE protocol and will give you things like this:
   
   Device Name (UUID 0x2A00): Shine
   
   Appearance (UUID 0x2A01): [1088] Generic: Running Walking Sensor
   
   Firmware Revision String (UUID 0x2A26): 0.0.50r
   
   Not very interesting. You can also get your device serial number and stuff like that. There’s another service on the device, and it’s a proprietary service with its own UUID that I won’t bother typing here. It has writable characteristics, but, without knowing what I’m doing, I’m reluctant to try writing to them. They may not even be related to the primary function of the device.
   
   I guess some things will remain a black box. I threw the device back together, and I’m going wear it for a while to see how it holds up! After all, even if you can’t hack it, Misfit has hinted at the release of an API in the future, and we can all look forward to that.

    

   jfieb, 17 minutes agoEditDeleteReport

   #2Reply
3. 

   jfiebMember

   Runtastic, Misfit, start to get the feel that more and more will happen just like this.
   
   
   
   Misfit launches a development kit that gives other companies access to its sensor algorithms
   Kif Leswing
   Sep. 4, 2014 - 10:00 AM PST
   
   Misfit, a company best known for making the Shine fitness tracker, has been in the process of transitioning to a company that sells its sleep and activity tracking algorithms in addition to tasteful hardware. On Thursday, it took that process one step farther and announced the Misfit Developer Toolkit, which will allow interested developers and companies to incorporate Misfit’s activity tracking capabilities into their own applications.
   
   There will be three elements to the Misfit toolkit. There will be a cloud storage API, which will provide the hooks needed to sync Misfit data between devices and third party applications. There’s also a Device SDK, which gives app makers the tools they need to interact with Misfit hardware. But perhaps the most interesting part of the new developer toolkit is the “Misfit Scientific Library,” which will give both hardware and software makers access to Misfit’s proprietary sensor algorithms.
   
   
   In addition to RunKeeper and MapMyFitness, which were previously announced Misfit partners, there are a few other big companies that will be taking advantage of possibilities afforded by the new SDK. Yesterday,Misfit announced a partnership with IFTTT that allows users to automate tracking with the Shine, such as automatically logging sleep into a Google spreadsheet.
   
   While Misfit designed and distributed the Shine, other more recent Misfit-branded hardware, like the Beddit Sleep System, merely integrates with and takes advantage of Misfit’s sleep tracking algorithms. Another example: Even though the Shine can be worn on the wrist, Misfit went ahead and released an app for the Pebble that tracks activity without the Shine. An SDK is a big step on this road for Misfit, which is angling to be the software glue that connects various different wearable products, even if Misfit’s not making them itself.
   
   You can download the Misfit developer’s kit here. If you’re interested in hearing more about the design of Misfit’s Shine device, come check out our experience design conference Roadmap, where the Shine’s designer, Mladen Barbaric, will be speaking

    

   jfieb, 11 minutes agoEditDeleteReport

   #3Reply
4. 

   jfiebMember

   Misfit CEO: Two new Misfit wearables launching at CES and only one is a tracker
   We chat trackers, China and smart homes with Sonny Vu
   
   Friday
   November 6, 2015
   By Sophie Charara
   @sophiecharara
   Now that Sonny Vu, Misfit's CEO, has built what he thinks is as close to the perfect activity, sleep and lifestyle tracker in the Shine 2, his attentions are moving elsewhere.
   
   "Activity tracking is great but it's very 2013," he said at the Shine 2's UK launch in London. "The era of wearable controls is here. Controls, identity, payments, we're on it. We have a lot of software updates coming in the next six months."
   
   And that's not all. Misfit's prolific run of launching new devices and collaborations looks set to continue in 2016.
   
   "We are launching two new wearables at CES and only one of them is an activity tracker," he told Wareable. "CES is going to be big for us, we are also announcing a new brand partnership there. But I don't know that we're going to do anything with a display soon. There just isn't a display that is low power enough."
   
   Misfit activity and sleep tracking is also "coming soon" to Android Wear watches, a move which would help Vu's claims in 2014 that Misfit is a platform, not just a devices company.
   
   Getting the details right
   
   
   When the Misfit Shine 2 was announced, we described it as a "fine-tuned tracker". And in many ways it is.
   
   According to Vu, people love the Shine's form factor and functionality but the perfectionists in his team wanted to finish the job before moving on to new products.
   
   "We were focused on getting all of the details right that we didn't get right for Shine," he said. "People still buy it, they like the iconic design, but this takes the device to the next level."
   
   The Shine 2 is 10% thinner and adds coloured lights and vibrations for alerts with a motor that's ten times lower power than smart jewellery rivals. Plus it brings Link smart home and selfie controls to a classy-looking wearable as well as Misfit's cheaper buttons. There's a redesigned app and a smart home-friendly 10 - 15m range.
   
   Read this: Eric Migicovsky explains why Pebble is in it for the long haul
   
   But Misfit is also getting the basics right. A new capacitive touch sensor means the double tap motion is quicker and more accurate, a niggle Shine users will be familiar with. It has made sure, for instance, that the new Shine doesn't fall out of the bands so easily, a common complaint all over its Amazon reviews and as Vu says, "kind of a basic requirement". A new necklace accessory is also in the pipeline.
   
   Misfit decided to shun San Francisco to launch the Shine 2 at the W Hotel in Beijing in late October, in collaboration with China's answer to both Amazon and Kickstarter, JD.com. It sold just under 250,000 units in pre-orders in the first 24 hours which Vu says is "more than Fitbit and Jawbone combined" for recent launch day sales.
   
   "We've never seen anything like it, there is a lot of momentum in China," said Vu. "Americans are more function over form, broadly speaking, you go to New York it's different. Whereas you have places like Italy and France, if you read our Amazon reviews, people love the design, we are on like 4 ½ stars. They just love the design. In countries that care about quality and design, we do better so in Japan we do really well and we do pretty well in the UK."
   
   
   
   Earlier this year, Vu told the press that Misfit was outselling Fitbit in China. The Swarovski Shine collaboration has also been seriously popular in both China and the US even though it is more expensive than most rivals.
   
   "The Americans love it because it feels more elegant, more European and Asians absolutely love it. A company like Polar will buy crystals and glue them on. That's not a Swarovski branded product, they're selling in their stores. The crystal quality is great and it sells really well."
   
   Read this: #Trending - bling boosted wearables
   
   As for whether we will be able to buy the solar powered Swarovski Shine any time soon, this is "going to be up to Swarovski".
   
   Misfit also claims to have translated its app into more languages than any other competitor. Whichever country Vu is selling in, he is betting on the combination of an elegant product, smartwatch-style features and no charging.
   
   "This does everything you would expect a smartwatch to do," he said. "It tells the time as a minimal watch, it handles alerts, it tracks activity and sleep, has a smart alarm. But it's the no charging that wins. It's 30 times longer battery life, even smart jewellery you have to charge it. You don't charge your ring now. Touchscreens have a place but Misfit is 24/7, 365."

    

   jfieb, 7 minutes agoEditDeleteReport

   #4Reply
5. 
   

   jfiebMember

   Understand that in the coral reef of wearable sensor fusion, everybody knows everybody, and they talk about what they need to do/have done.
   
   Look Dr Saxe in March….
   
   
   Mar 4, 2015
   Previous Release | Next Release 
   PDF
   Add to Briefcase
   Media Alert: QuickLogic to Present and Demonstrate at the Wearable Technology Show in London
   
   
   SUNNYVALE, CA -- (Marketwired) -- 03/04/15 -- QuickLogic Corporation (NASDAQ: QUIK), the innovator of ultra-low power programmable sensor hubs, will be presenting, participating in a panel and demonstrating its latest sensor hub and mobile technology solutions at the Wearable Technology Show in ExCel Conference Center, London.
   
   Demonstration:
   Stand 216, Fitness and I-Health Zone
   March 10 and 11, 2015 from 8:30 a.m. - 5:30 p.m.
   
   Panel: How will the promise of the "always on" society be realized? Can wearables and connected devices work together?
   Moderator: Nick Spencer, Senior Analyst, ABI
   Panelists: Dr. Tim Saxe will be joined by James Stables, Editor, Wareable, Greg Jones, Director International Business, Misfit, and Francois Jeanneau, CEO, Novasentis.
   March 11 at 10:20 a.m.
   
   Presentation: The Design Challenges of Wearable Technologies: How to Pack Performance, Power and Advanced Capabilities into a Small Form Factor Device
   Speaker: Dr. Tim Saxe, CTO QuickLogic
   March 11 at 12:30 p.m.
   Like RUNTASTIC some of the current projects will get moved up into bigger ecosystems just like that;
   
   Could it have already happened again?
   
   Maybe, but if not another one just like this, or maybe even better.
   But CES might be more fun than most imagine?