OctaveTutorialAndrewNg Andrew Ng Machine learning course Octave Tutorial transcripts .. (pdf)
I am looking forward to attending two almost sold out events – The tricky first million and scaling startups
If you are there, happy to meet
The Tricky First Million
26 March 14:00 – 20:00
“Your first million revenue or users are the hardest”
The importance of vision: act big from the start
MVPs & staying customer focussed
Selling your dream: to clients, recruits and staff
Raising money: traction, vanity metrics, how to choose an investor
When to hire key team members
Building the right culture
27 March 08:30 – 20:00
Building a multi-million pound business
Product quality as a driver of growth
Excellence in Recruiting
Managing company communication and culture
Creating processes that maintain agility and innovation
KPIs & data driven decision-making
How the role of the founder changes with growth
As you know from previous posts, I have been very interested in IoT / Smart cities and Algorithms
So, its nice to conduct this workshop based on the program ”Big data analytics and algorithms for cities” at the City sciences program for the Technical University of Madrid
IoT and Machine Learning is a unique one day workshop which explores Machine learning techniques for IOT.
The workshop is designed as an exploratory/introductory workshop for participants who are interested in using machine learning techniques for IoT data.
Arthur Samuel, the pioneeing AI scientist, defined machine learning as – ‘The field of study that gives computers the ability to learn without being explicitly programmed’. Machine learning includes examples such as the Driverless car which require data from other cars, street lights, people and a range of sensors coupled with the analytics to make real-time decisions.
Hence, unlike programmers who work with pre-defined logic for a problem domain( using statements like if-then-else, loops etc), for data scientistics, the logic is often non-deterministic.
Thus, given an IOT data set, the machine learning algorithm has to deduce a logic based on a pattern in the data.
The first part of the workshop will explain Machine learning techniques. This will be followed by understanding how these techniques could be applied to IOT datasets. We will use Smart city datasets to explore Machine learning and we will explore specific techniques like sensor fusion.
Machine learning techniques we explore are:
- Supervised and unsupervised learning
- Neural Networks
- Machine Learning System Design
- Anomaly Detection
- Recommender Systems
- Large-Scale Machine learning systems
- Programming paradigms and Languages for machine learning
- Computation at the edge or Computation at the core
Problem domains include:
- Prediction Examples bassed on datasets (energy, pollution)
- Optimzation based (traffic routing, commute optimization)
- Pattern identifying (predict hotspots based on health care data)
- New business proceeses based on machine learning for objects that have to navigate an unpredictable domain (driverless cars, drones)
Note that this course is introductory but still needs basic understand and aptitude for Mathematics
Please contact me for any queries/thoughts/comments
workshop link in Palo Alto on IoT and Machine Learning
Here is my review of the Mobile World Congress. This year, for the first time, I was invited to attend the GSMA Ministerial programme as part of my work in being the co-author for Digital world in 2030 report to be released in the European parliament next week
This blog is a personal perspective (i.e. related to the above report – but my own views).
It tracks disruptive trends and I present a perspective / viewpoint on the industry. It is biased towards IOT – which is also a personal focus – especially due to my teaching for ”Big data analytics and algorithms for cities” at the City sciences program for the Technical University of Madrid
We now live in a mobile first world. I was not carrying a laptop and this single image (Android with Intel inside) – shows us that we are indeed in a Mobile first world
The ‘Mobile first world’ also explains the sheer size of MWC(85,000 attendees – a 20% jump from the previous year) – 45 restaurants – two heliports – 1,700 exhibitors – eight exhibition halls – 240,000 square meters (2.6 million square feet)
If you are not a part of it – you are missing out!
So, here are my perspectives ..
Firstly, here is my overall perspective, which I also discussed in my keynote at the Swiss mobile association in Zurich (slides) (smama)
- The Mobile data industry as it stands today, is about fifteen years old.
- It is fast growing but mature. We now have a two horse race for devices (considering Samsung owns about 70% of the Android market).
- In innovation terms, innovation shifts from building networks to creating light bulbs. When compared to electric networks – in the early stages – there was a period in which networking technologies competed (AC/DC – Edison/Tesla) etc. Once that was decided, innovation shifts to creating light bulbs). We are in a similar phase.
- So, innovation shifts along three dimensions: Horizontal (apps), Vertical (cross stack – ex IOT) and network.
- Why is network the third dimension of innovation? Unlike electric and power networks which have a 50/60 year cycle – Telco networks have a 7 year cycle (from standardization to spectrum to devices). That means the rate of Telco innovation is also comparatively faster and every network innovation leads to a knock on effect for secondary innovation (ex devices). Hence, 5G is important What will 5G look like and how 5G will shape the technological landscape of countries for the next decade
- Since 5G is widely expected to be deployed around 2020 – the question then is: How will the market play out between now and 2020? There are some key indicators already: iBeacon could be a de-facto standard since it’s an ‘open enough’ standard. (I.e. iPhone and Android devices can both use it). Many of the use cases for mobile couponing etc promoted by NFC – could well be deployed with iBeacon. More interestingly, iBeacon could motivate retailers to open up their WiFi networks and allow others to use their WiFi (in return for instore coupons)
- Similarly, Hotspot 2.0 would allow us to seamlessly navigate between cellular and wifi networks.
- This means – by 2020 – we could live in a world with primarily localized connectivity and 5G would then make that connectivity pervasive (like ‘air’)
- In this world, Operators would have a much better visibility of their customers(through say Hotspot 2.0) and would focus on being much more customer centric by truly leveraging their data (Big Data, IoT etc)
- To quote John F Kennedy, a rising tide will lift all boats and thus, we have an optimistic view of the industry
- Thus, Hotspot 2.0, Low energy Bluetooth (on which iBeacon is based) will be interesting. It will be also interesting to see how Zigbee fares in future in context of Bluetooth (and the jury is definitely out on Zigbee as is on NFC (Apple’s ibeacon mobile payment system is the death knell of NFC). NFC and Zigbee have taken too long to gain critical mass – and normally when that happens – something else takes over
- More interesting from a connectivity standpoint is 900Mhz wifi also called 80211ah standard. At MWC 900MhzWi-Fi makes a debut which could provide wi-fi with dedicated bands for ultra-reliable, always-on machine-to-machine connections.
Why the focus on IOT ..
2014 was clearly the year of IOT. IOT is not M2M (machine to machine). M2M is a telecoms term which implies that there is a radio (cellular) at both ends of the communication. On the other hand, IOT means simply connecting to the Internet.
The two are not the same!
From an IOT perspective, we can see connectivity in three ways:
a) Things connected to the cellular network directly and communicating mainly via cellular (machine to machine)
b) Things connected to Internet, speaking to each other in an autonomous, discoverable, peer to peer mode (ubiquitous computing)
c) Things connected to the mobile phone and then to the Internet
(a) is too expensive for mass market. (b) is too futuristic. (c) is happening now .. and is my main focus here
MWC is all about devices!
- LG, Samsung, and Sony dominate ex Sony’s Xperia Z2
- In contrast, devices from Huawei and ZTE did not seem to be impressive. It will be interesting to see if they are successful going forward as device vendors or will continue to remain infrastructure vendors
- Both Nokia and Blackberry (former leaders) remain uncertain with Nokia adopting Android.
- New devices like Yotaphone 2 (screen on one side – e-reader on the other side) provide new interfaces
- Some vendors like Kyocera introduce some quirky concepts
- Both Sailfish (Jolla) and Ubuntu had a presence – but too early to say how these devices will fare in the future
- There were devices and then there was Samsung Galaxy S5! Apart from the usual (5.1 inch screen, 16 megapixels camera etc), we also have for the first time a fingerprint scanner and support for 128-GB memory card.
- The Galaxys5 also allows you to track your heartbeat. In conjunction with the Samsung Gear Fit , we have the makings of a new class of IOT device targeting healthcare
Emerging market devices
Devices targeting emerging markets and the cost conscious customer were prominent.
- Nokia X(with Android) strategy can be seen in this light since they support up to 75% of apps available on Google Play. The Nokia X, X+ and XL will cost €89 (£73), €99 (£81) and €109 (£89), respectively.
- The $25 firefox phone also comes in the same class and challenges the Nokia X
- The facebook whatsapp strategy could also been in that light
- In an age of Snowden, Privacy has taken center stage.
Mozilla announced the Future of Mobile Privacy to secure data easily.
- Samsung announced the Knox security product targeted at small and medium enterprises. According to Computerweekly “The SME offering allows dual “personalities” to be set up on a smartphone to allow one handset to function as two separate devices, keeping the data from each, whether personal or corporate, contained and secure.”
- And then there was the Blackphone – dubbed the Snowden phone (Blackphone web site). The ‘privacy-first’ runs a customized version Android OS.
- As mentioned before, the Galaxy 5 already has fingerprint recognition
- My analysis: It’s interesting to see how many people actually buy secure phones for themselves (i.e. not popular phones like Galaxy 5 which have secure features built into them). In my experience, many people (with Geographical exceptions like Germany) – want privacy – but actually behave in the opposite manner (or do not pay for privacy)
I have already explained the significance of IOT above .. Here are some interesting observations
- One of the stars of the show was the unexpected Oral B’s Smartseries toothbrush which sends data back to a smartphone app recording how many brushstrokes are used, targeting problem areas and following personalised brushing routines
- On the processor side freescale KL03 was announced and both Qualcomm and Intel launched Intel launched a 64 bit mobile processors.
- I got to see the Arduino compatible Intel Galileo for the first time. A more detailed blog post on this soon
- Cisco and SAP launched some interesting initiatives combining Cisco’s networking and the HANA platform ex Connected Vending, Connected Construction (Internet of Things at MWC – Cisco Blog)
- On the platform side, I follow Alljoyn (now the AllSeen Alliance) – Internet of Things consortium chair warns of Google’s Android effect
- Jasper wireless launched a Global SIM for M2M. Not sure if there will be a widespread uptake of this outside the Operator community
- Tizen! It was there .. In Gear2 and Gear 2 Neo Smart watches, Samsung has ditched Android in favour of Tizen. But for Gearfit – there is a proprietary Operating system. We also see no commercial smartphones running Tizen
- Big Data and analytics – The value for IOT will be in terms of data. Hardware platforms like Tessel and Espruino capture location. Plotly allows you to do data analysis and visualization based on data from Arduino, Tessel, and Raspberry Pi that is streamed to the Cloud.
- More exciting is techniques like sensor fusion. Sensor fusion is the generation of new, richer forms of actionable intelligence through the application of analytical processing techniques to diverse, multi-source data sets. Companies like Camgian are leaders in this space. (Three Trends That Will Accelerate IoT Enterprise Adoption).
- Smaller companies focussing on low energy Bluetooth(BeeWi)
- Outside of MWC, I follow Open source and IOT with interest including the following - open source thermostat competitor to NEST (Spark.io), Semantic web and IOT and web of things group at w3c
- I also spent time at the impressive Qualcomm/Allseen alliance demo
- and finally saw a demo of the IOT project from our friends at Red Ninja Studios in Liverpool and @ARMcommunity at the ARM booth
Comments/feedback welcome at ajit.jaokar at futuretext.com. Follow me @ajitjaokar
For a copy of the Digital world in 2030 report to be released in the European parliament next week – follow the link
What will 5G look like and how 5G will shape the technological landscape of countries for the next decade
Based on my observations at MWC and outside, here is an analysis of why 5G matters and how it could shape the technological landscape of countries.
What is 5G
The Ericsson white paper on 5G says “5G radio access is an integrated set of technologies addressing a wide variety of use cases and requirements”
This statement sounds very generic but covers a lot of specific details because technology does not exist in vacuum. Hence, 5G should be seen in context of the previous technological evolutions. These are: 1G(1980’s) Voice; 2G(1990’s) Voice & Text; 3G(2000’s) Multimedia; 4G(2010’s) Mobile Internet. Building on these, 5G could be seen as a set of technologies that bring the networked society to life. This statement ‘creating a set of technologies that bring the networked society to life’ also sounds aspirational ..
So we need to explore what 5G means for applications and services
Today, Mobile traffic is driven by predictable activities which include making calls, email communications, browsing, watching video etc. Over the next 5 to 10 years, it is expected that billions of new devices with less predictable traffic patterns will join the network. These include Smart meters, Automotive applications, remote surveillance etc.
So, to handle this, 5G represents a convergence of Radio access technologies. To this, 5G also adds a good user experience as a goal
What does 5G mean for applications and applications?
The drivers for 5G are
- Massive growth in Connected Devices
- Wide range of Requirements and Characteristics
- Making access affordable and sustainable
To achieve this in a diverse ecosystem as discussed above, from a network standpoint, we could classify applications along the parameters:
- high volume(classical)
- latency oriented
- data payloads
- energy efficiency
- network reliability
- Relaxed latency requirements – Remote meter reading for billing purposes
- Strict latency requirements – Safety or control mechanisms in the process industry,
- High network reliability – Applications relating to the control of critical infrastructure (such as electrical grids), industrial control etc
- High volume – Remote video surveillance,
- Low payload – Smart metering
And all these on a massive scale (50 billion)
5G will expand into new deployment scenarios, for example ultra-dense deployments where the distance between network access nodes may be as small as a few meters and to Direct device-to-device communication. Direct to device supports radio access between a terminal and an access point connected to the Internet. Direct to device is beneficial when devices that share information are close to each other and information has in a local context.
What to expect from next generation wireless networks technology?
Some of the qualities of a 5G network include:
- mobile data rates of multiple gigabits per second;
- higher spectral efficiency and advanced beam-forming technology;
- a dramatic reduction in power consumption by both devices and macro- and microcell sites;
- latencies of less than five milliseconds end to end;
- a larger number of supported devices;
- pervasive coverage;
- extreme reliability;
- self-healing capabilities.
The blog What to expect from the next generation of wireless communications? Gives a good outline of 5G features which I summarise below
End to end broadband connectivity that meets or exceeds user expectations – Higher data rates , superior end user experience, minimal dead zones and dropped data/voice packets, high quality of voice, ubiquitous Five signal bars, close to zero Latency, no network congestion, pervasive and robust networks that can withstand natural calamities
Dynamic allocation of resources –mapping the service requirements to the most suited combination of frequency and radio resources.
Customized user specific experience – The future wireless communications will focus increasingly on delivering enhanced support for diverse applications while remembering the customers’ preferences and usage patterns. The customer will be able to trust the service provider. Quality of service will be tailored to the applications’ demand.
Multi-hop mobile ad hoc networks – Mobile Ad Hoc Networks (MANETs) are designed to operate without any infrastructure and possess self-organizing capabilities. The wireless nodes communicate among themselves using multi-hop radio relaying, without the packets passing through a central access point or a base station.
The Internet of Things – The idea of machine to machine communications is not new and has been discussed for years. However, the ‘Internet of Things’ is a relatively fresh term referring to various objects in our lives all possessing a unique RFID (linked to an IP address). They will communicate with each other without human interference, help in decision making and save time.
Centimeter and millimeter wave frequencies as spectrum source –The spectrum above 3 GHz has been largely left untouched so far in most parts of the world. High frequencies means low range
Smart and efficient receivers –The primary intelligence of the network will remain with the transmitter for the foreseeable future, but eventually receivers will play an increasingly important role in making the mobile environment more efficient on a network level.
This is summarised in the following diagrams
The race for the next generation game changers is on ..
5G could be a game changer for countries and companies and many have realised that
The UK effort is led by the University of Surrey 5G research center
Ericsson led metis2020 project, the Intel Strategic Research Alliance for 5G, the EU project 5gnow , Huawei’s 5G collaborations in Europe (especially in Munich) and ofcourse Neelie Kroes announcements for 5G are examples of various projects in this space
The World Radiocommunications Conference, scheduled for November 2015. Will bring many of these discussions into focus
The new 5G-Public Private Partnership will work in three phases over the next six years to define and launch the 5G infrastructure:
- Phase I (2014-2016): Basic research work / Vision building
- Phase II (2016-2018): System optimization / Pre-standardization
- Phase III (2018-2020): Large-scale trials / Early Standardization
To conclude, the discussions for 5G are only just beginning! And much is at stake
A unified view of a 5G network would look as below
Its always nice to speak in German speaking countries since I have so many friends and also family in Germany
Zurich has been a hotbed of mobile activity with Google’s labs there – so its great to speak in Zurich for this event
the next trade event will be held on 6 . March 2014 in the HWZ an exciting topic ‘mTrends’ instead. Please save the date!
Are you interested? Your application we will be happy to take on 3 March against 2014. All other details can be found below.
Participation is for smama members (max. 3 persons per company + max. 3 guests) free. Non-members and additional guests pay a fee of CHF 80 -.
Fritz Reust, Managing smama
16:30 to 17:00 Arrival of the guests
17:00 – Welcome by Fritz Reust, Managing smama
17:00 to 17:30 Keynote 1: “The Mobile World in 2018: The future is based on global micro-trends” - Sebastian Rassmann,Trendone, Hamburg
17:30 to 18:45 Top 5 Trends for 2014
The smama departmental managers designate the mTrends 2014:
- Ronie guarantor, mobagogo to mMarketing: Couponing – LBS / LBA – CRM – Offline to Online Cross Chanel. mAdvertising: New advertising formats – Optimization – BigData – Local Based Advertising – programatic Buying
- Joachim Hagger, Netcetera to mApps & mInternet: cross-channel services – Real-time Web – mobilization of Business – Indoor navigation – Wearables – extinction of the desktop Web
- Tobias Wirth, Aduno group to mCommerce & mPayment: Mobile Contactless Payment – M-wallets – Beacon / iBeacon – In-Store Remote M-Payment – P2P Purse to Purse M-Payment
- Martin Fessler, view group to mContent & mPublishing: Personalization – Mobile First – Advertising – Crowd
- Sascha Corti, Microsoft to MTechnology: Device Trends – Internet of Things – BYOD within the company – Enterprise Apps – HTML5/CSS/JS as native development language
- Dr. Oliver Staffelbach, Wenger & Vieli to mStartups: developments in Startups – Presentation Sharely
18:45 to 19:15 Keynote 2: “Cross Domain technologytrends in Telecoms, Mobile and the Internet” - Ajit Jaokar,opengardensblog / Future Text, London
19:15 to 19:30 Q & A
19:00 to 20:30 Aperitif / Networking
21:00 end of the event
Here you will find the gesame program. ((Link follows))
I am back from MWC and also from the keynote at the Raspberry Pi conference. More on those soon.
Independent of MWC, here is an interesting link I saw about 5G from Stephen Temple’s blog
The question is: Will 5G benefit all (valleys) Or will it benefit a few who already have good coverage (hills)
Stephen identifies three trends affecting Spectrum allocation
a) Radio system technology advances are focused solely on making the hills ever higher but doing relatively little about the valleys.
b) Spectrum regulators are doing their best to find more spectrum to meet the rising consumer demand for wireless data (again – the hills)
c) The third trend is the collapse of mobile handset performance as an ever greater number of frequency bands are brought into use for mobile radio. An antenna optimised to work over widely separated frequency bands always has a worse performance than one optimised for a specific frequency band.
You can read the full story from the blog - Will 5G be for a minority of high capacity hill dwellers or for the rest of us?
Image source Stephen Temple blog representing Hills and Valleys
Note – Overall, I like what Linkedin is doing including the new publishing platform . I also ‘trust’ Linkedin to upload all my contacts because its a great networking feature. I only have a problem with the specific issue (as a father) as below
Like many people, I uploaded my contacts on Linkedin
Its a useful feature
I was surprised to find my son’s name (and email address) as a Linkedin profile as ‘People You May know’
My son is aged 10 ..
And is very tech savvy .. Here is a book we are writing together – Python in Schools – A Computer Science perspective – By Ajit Jaokar and Aditya Jaokar
But as far as I know, he does not have a Linkedin profile
At first, I thought he had created one
The poor child denied it ..
And then I realized that Linkedin is creating dummy profiles from all emails I unload – whether they have a existing Linked in profile or not
Why supposedly respected / corporate sites should use such tacky techniques is beyond me!
And then companies complain when they are under the cross fire of EU regulation
Who ELSE can see my son’s email address I don’t know ..
But behavior like this deserves to be highlighted because its a lack of trust
If minor’s email addresses are visible to others – it could have serious legal and regulatory implications for Linkedin
If anyone in Linkedin writes back I shall update this
See also this post – which I suspect is the same issue.
Update from Linkedin
Linkedin has formally contacted me about this. Their response is: They do not store profiles. However, they have no way of being able to distinguish an email address i.e. when they upload an address book, they upload all contacts (including in this case, my son). Apparently, this cannot be avoided.
Its good Linkedin does not expose this email address to others. But still sounds very opportunistic. After all, this is the company that has invented ‘Big Data’ and email addresses can be hacked(as has happened to Linkedin itself). So, all in all, it does not seem to be that bad – but could be improved none the less
I am at Mobile World Congress and this year for the first time, I am invited to the Ministerial program at Mobile World Congress which deals with Technology policy issues
This is due to my work as co-author for the European Internet Foundation – Digital world in 2030 program
The High level roundtable for the report is on Sunday which I am also attending