Guest blog post by Cameron Turner

Executive Summary

Though often the focus of the urban noise debate, Caltrain is one of many contributors to overall sound levels along the Bay Area’s peninsula corridor. In this investigation, Cameron Turner of Palo Alto’s The Data Guild takes a look at this topic using a custom-built Internet of Things (IoT) sensor atop the Helium networking platform.

Introduction

If you live in (or visit) the Bay Area, chances are you have experience with the Caltrain. Caltrain is a commuter line which travels 77.4 miles between San Francisco and San Jose , carrying over 50 thousand passengers on over 70 trains daily.[1]

I’m lucky to live two blocks from the Caltrain line, and enjoy the convenience of the train. My office, The Data Guild, is just one block away. The Caltrain and its rhythms, bells and horns are a part of our daily life, and connect us to the City and with connections to BART, Amtrak, SFO and SJC, the rest of the world.

Over the holidays, my 4-year-old daughter and I undertook a project to quantify the Caltrain through a custom-built sensor and reporting framework, to get some first-hand experience in the so-called Internet of Things (IoT). This project also aligns with The Data Guild’s broader ambition to build out custom sensor systems atop network technologies to address global issues. (More on this here.)

Let me note here that this project was an exploration, and was not conducted in a manner (in goals or methodology) to provide fodder for either side of the many ongoing caltrain debates: the electrification project, quiet zone, or tragic recent deaths on the tracks.

Background

My interest in such a project began with an article published in the Palo Alto Daily in October 2014. The article addressed the call for a quiet zone in downtown Palo Alto, following complaints from residents of buildings closest to the tracks. Many subjective frustrations were made by residents based on personal experience.

According the the Federal Railroad Administration (FRA), the rules by which Caltrain operates, train engineers “must begin to sound train horns at least 15 seconds, and no more than 20 seconds, in advance of all public grade crossings.”

Additionally: “Train horns must be sounded in a standardized pattern of 2 long, 1 short and 1 long blasts.” and “The maximum volume level for the train horn is 110 decibels which is a new requirement. The minimum sound level remains 96 decibels.“

Questions

Given the numeric nature of the rules, and the subjective nature of current analysis/discussion, it seemed an ideal problem to address with data. Some of the questions we hoped to address including and beyond this issue:

• Timing: Are train horns sounded at the appropriate time?
• Schedule: Are Caltrains coming and going on time?
• Volume: Are the Caltrain horns sounding at the appropriate level?
• Relativity: How do Caltrain horns contribute to overall urban noise levels?

Methodology

Our methodology to address these topics included several steps:

1. Build a custom sensor equipped to capture ambient noise levels
2. Leverage an uplink capability to receive data from the sensor in near real-time
3. Deploy sensor then monitor sensor output and test/modify as needed
4. Develop a crude statistical model to convert sensor levels (voltage) to sound levels (dB)
5. Analysis and reporting

Apparatus

We developed a simple sensor based on the Arduino platform. A baseline Uno board, equipped with a local ATmega328 processor, was wired to and Adafruit Electret Microphone/Amplifier 4466 w/adjustable gain.

We were lucky to be introduced through the O’Reilly Strata NY event to a local company: Helium. Backed by Khosla Ventures et al, Helium is building an internet of things platform for smart machines. They combine a wireless protocol optimized for device and sensor data with cloud-based tooling for working with the data and building applications.

We received a Beta Kit which included a Arduino shield for uplink to their bridge device, which then connects via GSM to the Internet. Here is our sensor (left) with the Helium bridge device (right).

Deployment

With our instrument ready for deployment, we sought to find a safe location to deploy. By good fortune, a family friend (and member of the staff of the Stanford Statistics department, where I am completing my degree) owns a home immediately adjacent to a Caltrain crossing, where Caltrain operators are required to sound their horn.

Conductors might also be particularly sensitive to this crossing, Churchill St., due to its proximity to Palo Alto High School and the tragic train-related death of a teen, recently.

From a data standpoint, this location was ideal as it sits approximately half-way between the Palo Alto and California Avenue stations.

We deployed our sensor outdoors facing the track in a waterproof enclosure and watched the first data arrive.

Monitoring

Through a connector to Helium’s fusion platform, we were able to see data in near real-time. (note the “debug” window on the right, where microphone output level arrives each second).

We used another great service, provided by Librato, (now a part of SolarWinds) a San Francisco-based monitoring and metrics company. Using Librato, we enabled data visualization of the sound levels as they were generated. We were able to view this relative to its history. This was a powerful capability as we worked to fine-tune the power and amplifier.

Note the spike in the middle of the image above, which we could map to a train horn heard ourselves during the training period.

Data Preparation

Next, we took a weekday (January 7, 2015), which appeared typical of a non-holiday weekday relative to the entire month of data collected. For this period, we were able to construct a 24-hour data set at 1-second sample intervals for our analysis.

Data was accessed through the Librato API, downloaded as JSON, converted to CSV and cleansed.

Analysis

First, to gain intuition, we took a sample recording gathered at the sensor site of a typical train horn.

Click HERE to hear the sample sound.

Using matplotlib within an ipython notebook, we are able to “see” this sound, in both its raw audio form and as a spectrogram showing frequency:

Next, we look at our entire 24 hours of data, beginning on the evening of January 6, and concluding 24 hours later on the evening of January 7th. Note the quiet “overnight” period, about a quarter of the way across the x axis.

To put this into context, we overlay the Caltrain schedule. Given the sensor sits between the Palo Alto and California Avenue stations, and given the variance in stop times, we mark northbound trains using the scheduled stop at Palo Alto (red), and southbound trains using the scheduled stop at California Ave (green).

Initially, we can make two converse observations: many peak sound events tend to lie quite close to these stop times, as expected. However: many of the sound events (including the maximum recorded value, the nightly ~11pm freight train service) occur independent of the scheduled Caltrains.

Conversion to Decibels

On the Y axis above, the sound level is reported in the raw voltage output from the Microphone. To address the questions above we needed a way to convert these values to decibel units (dB).

To do so, a low-cost sound meter was obtained from Fry’s. Then an on-site calibration was performed to map decibel readings from the sensor to the voltage output uploaded from our microphone.

Within R Studio, these values were plotted and a crude estimation function was derived to create a linear mapping between voltage and dB:

The goal of doing a straight line estimate vs. log-linear was to compensate for differences in apparatus (dB meter vs. microphone within casing) and overall to maintain conservative approximations. Most of the events in question during the observation period were between 2.0 and 2.5 volts, where we collected several training points (above).

A challenge in this process was the slight lag between readings and data collection with unknown variance. As such, only “peak” and “trough” measurements could be used reliably to build the model.

With this crude conversion estimator in hand, we would now replot the data above with decibels on the y axis.

Clearly the “peaks” above are of interest as outliers from the baseline noise level at this site. In fact, there are 69 peaks (>82 dB) observed (at 1-second sample rate), and 71 scheduled trains for this same period. Though this location was about 100 yards removed from the tracks, the horns are quieter than the recommended 96dB-115dB range recommended by the FRA. (With caveat above re: crude approximator)

Interesting also that we’re not observing the “two long-two short-one long” pattern. Though some events are lost to the sampling rate, qualitatively this does not seem to be a standard practice followed by the engineers. Those who live in Palo Alto also know this to be true, qualitatively.

Also worth noting is the high variance of ambient noise, the central horizontal blue “cloud” above, ranging from ~45 dB to ~75 dB. We sought to understand the nature of this variance and whether it contained structure.

Looking more closely at just a few minutes of data during the Jan 7 morning commute, we can see that indeed there is a periodic structure to the variance.

In comparing to on-site observations, we could determine that this period was defined by the traffic signal which sits between the sensor and the train tracks, on Alma St. Additionally, we often observe an “M” structure (bimodal peak) indicating the southbound traffic accelerating from the stop line when the light turned green, followed by the passing northbound traffic seconds later.

Looking at a few minutes of the same morning commute, we can clearly see when the train passed and sounded its horn. Here again, green indicates a southbound train, red indicates and northbound train.

In this case, the southbound train passed slightly before its scheduled arrival time at the California Avenue station, and the Northbound train passed within its scheduled arrival minute, both on time. Note also the peak unassociated with the train. We’ll discuss this next.

Perhaps a more useful summary of the data collected is shown as a histogram, where the decibels are shown on the X axis and the frequency (count) is shown on the Y axis.

We can clearly see a bimodal distribution, where sound is roughly normally distributed, with a second distribution at the higher end. The question still remained why several of the peak observed values fell nowhere near the scheduled train time?

The answer here requires no sensors: airplanes, sirens and freight trains are frequent noise sources in Palo Alto. These factors, coupled with a nearby residential construction project accounted for the non-regular noise events we observed.

Click HERE to hear a sample sound.

Finally, we subsetted the data into three groups, one to look at non-Train minutes, one to look at northbound train minutes and one to look at southbound train minutes. The mean dB levels were 52.13, 52.18 and 52.32 respectively. While the order here makes sense, these samples bury the outcome since a horn blast may only be one second of a train-minute. The difference between northbound and southbound are consistent with on-site observation-- given the sensor lies on the northeast corner of the crossing, horn blasts from southbound trains were more pronounced.

Conclusion

Before making any conclusions it should be noted again that these are not scientific findings, but rather an attempt to add some rigor to the discussion around Caltrain and noise pollution. Further study with a longer period of analysis and duplicity of data collection would be required to statistically state these conclusions.

That said, we can readdress the topics in question:

Timing: Are train horns sounded at the appropriate time?

The FRA recommends engineers sound their horn between 15 and 20 seconds before a crossing. Given the tight urban nature of this crossing this recommendation seems a misfit. Caltrain engineers are sounding within 2-3 seconds of the crossing, which seems more appropriate.

Schedule: Are Caltrains coming and going on time?

Though not explored in depth here, generally we can observe that trains are passing our sensor prior to their scheduled arrival at the upcoming station.

Volume: Are the Caltrain horns sounding at the appropriate level?

As discussed above, the apparent dB level at a location very close to the track was well below the FRA recommended levels.

Relativity: How do Caltrain horns contribute to overall urban noise levels?

The Caltrain horns generate roughly an additional 10dB to peak baseline noise levels, including period traffic events at the intersection observed.

Opinions

Due to their regular frequency and physical presence, trains are an easy target when it comes to urban sound attenuation efforts. However, the regular oscillations of traffic, sirens, airplanes and construction create a very high, if not predictable baseline above which trains must be heard.

Considering the importance of safety to this system, which operates just inches from bikers, drivers and pedestrians, there is a tradeoff to be made between supporting quiet zone initiatives and the capability of speeding trains to be heard.

In Palo Alto, as we move into an era of electric cars, improved bike systems and increased pedestrian access, the oscillations of noise created by non-train activities may indeed subside over time. And this in turn, might provide an opportunity to lower the “alert sounds” such as sirens and train horns required to deliver these services safely. Someday much of our everyday activity might be accomplished quietly.

Until then, we can only appreciate these sounds which must rise above our noisy baseline, as a reminder of our connectedness to the greater bay area through our shared focus on safety and convenient public transportation.

…

Acknowledgements:

Sincere thanks to Helen T. and Nick Parlante of Stanford University, Mark Phillips of Helium and Nik Wekwerth/Jason Derrett/Peter Haggerty of Librato for their help and technical support.

Thanks also to my peers at The Data Guild, Aman, Chris, Dave and Sandy and the Palo Alto Police IT department for their feedback.

And thanks to my daughter Tallulah for her help soldering and moral support.

[1] http://en.wikipedia.org/wiki/Caltrain

Originally posted on LinkedIn. 

Follow us @IoTCtrl | Join our Community

Guest blog post by ajit jaokar

By Ajit Jaokar @ajitjaokar Please connect with me if you want to stay in touch on linkedin and for future updates

Cross posted from my blog - I look forward to discussion/feedback here

Note: The paper below is best read as a pdf which you can download from the blog for free

Background and Abstract

This article is a part of an evolving theme. Here, I explain the basics of Deep Learning and how Deep learning algorithms could apply to IoT and Smart city domains. Specifically, as I discuss below, I am interested in complementing Deep learning algorithms using IoT datasets. I elaborate these ideas in the Data Science for Internet of Things program which enables you to work towards being a Data Scientist for the Internet of Things  (modelled on the course I teach at Oxford University and UPM – Madrid). I will also present these ideas at the International conference on City Sciences at Tongji University in Shanghai  and the Data Science for IoT workshop at the Iotworld event in San Francisco

Please connect with me if you want to stay in touch on linkedin and for future updates

Deep Learning

Deep learning is often thought of as a set of algorithms that ‘mimics the brain’. A more accurate description would be an algorithm that ‘learns in layers’. Deep learning involves learning through layers which allows a computer to build a hierarchy of complex concepts out of simpler concepts.

The obscure world of deep learning algorithms came into public limelight when Google researchers fed 10 million random, unlabeled images from YouTube into their experimental Deep Learning system. They then instructed the system to recognize the basic elements of a picture and how these elements fit together. The system comprising 16,000 CPUs was able to identify images that shared similar characteristics (such as images of Cats). This canonical experiment showed the potential of Deep learning algorithms. Deep learning algorithms apply to many areas including Computer Vision, Image recognition, pattern recognition, speech recognition, behaviour recognition etc

How does a Computer Learn?

To understand the significance of Deep Learning algorithms, it’s important to understand how Computers think and learn. Since the early days, researchers have attempted to create computers that think. Until recently, this effort has been rules based adopting a ‘top down’ approach. The Top-down approach involved writing enough rules for all possible circumstances.  But this approach is obviously limited by the number of rules and by its finite rules base.

To overcome these limitations, a bottom-up approach was proposed. The idea here is to learn from experience. The experience was provided by ‘labelled data’. Labelled data is fed to a system and the system is trained based on the responses. This approach works for applications like Spam filtering. However, most data (pictures, video feeds, sounds, etc.) is not labelled and if it is, it’s not labelled well.

The other issue is in handling problem domains which are not finite. For example, the problem domain in chess is complex but finite because there are a finite number of primitives (32 chess pieces)  and a finite set of allowable actions(on 64 squares).  But in real life, at any instant, we have potentially a large number or infinite alternatives. The problem domain is thus very large.

A problem like playing chess can be ‘described’ to a computer by a set of formal rules.  In contrast, many real world problems are easily understood by people (intuitive) but not easy to describe (represent) to a Computer (unlike Chess). Examples of such intuitive problems include recognizing words or faces in an image. Such problems are hard to describe to a Computer because the problem domain is not finite. Thus, the problem description suffers from the curse of dimensionality i.e. when the number of dimensions increase, the volume of the space increases so fast that the available data becomes sparse. Computers cannot be trained on sparse data. Such scenarios are not easy to describe because there is not enough data to adequately represent combinations represented by the dimensions. Nevertheless, such ‘infinite choice’ problems are common in daily life.

How do Deep learning algorithms learn?

Deep learning is involved with ‘hard/intuitive’ problem which have little/no rules and high dimensionality. Here, the system must learn to cope with unforeseen circumstances without knowing the Rules in advance. Many existing systems like Siri’s speech recognition and Facebook’s face recognition work on these principles.  Deep learning systems are possible to implement now because of three reasons: High CPU power, Better Algorithms and the availability of more data. Over the next few years, these factors will lead to more applications of Deep learning systems.

Deep Learning algorithms are modelled on the workings of the Brain. The Brain may be thought of as a massively parallel analog computer which contains about 10^10 simple processors (neurons) – each of which require a few milliseconds to respond to input. To model the workings of the brain, in theory, each neuron could be designed as a small electronic device which has a transfer function similar to a biological neuron. We could then connect each neuron to many other neurons to imitate the workings of the Brain. In practise,  it turns out that this model is not easy to implement and is difficult to train.

So, we make some simplifications in the model mimicking the brain. The resultant neural network is called “feed-forward back-propagation network”.  The simplifications/constraints are: We change the connectivity between the neurons so that they are in distinct layers. Each neuron in one layer is connected to every neuron in the next layer. Signals flow in only one direction. And finally, we simplify the neuron design to ‘fire’ based on simple, weight driven inputs from other neurons. Such a simplified network (feed-forward neural network model) is more practical to build and use.

Thus:

a)      Each neuron receives a signal from the neurons in the previous layer

b)      Each of those signals is multiplied by a weight value.

c)      The weighted inputs are summed, and passed through a limiting function which scales the output to a fixed range of values.

d)      The output of the limiter is then broadcast to all of the neurons in the next layer.

Image and parts of description in this section adapted from : Seattle robotics site

The most common learning algorithm for artificial neural networks is called Back Propagation (BP) which stands for “backward propagation of errors”. To use the neural network, we apply the input values to the first layer, allow the signals to propagate through the network and read the output. A BP network learns by example i.e. we must provide a learning set that consists of some input examples and the known correct output for each case. So, we use these input-output examples to show the network what type of behaviour is expected. The BP algorithm allows the network to adapt by adjusting the weights by propagating the error value backwards through the network. Each link between neurons has a unique weighting value. The ‘intelligence’ of the network lies in the values of the weights. With each iteration of the errors flowing backwards, the weights are adjusted. The whole process is repeated for each of the example cases. Thus, to detect an Object, Programmers would train a neural network by rapidly sending across many digitized versions of data (for example, images)  containing those objects. If the network did not accurately recognize a particular pattern,  the weights would be adjusted. The eventual goal of this training is to get the network to consistently recognize the patterns that we recognize (ex Cats).

How does Deep Learning help to solve the intuitive problem

The whole objective of Deep Learning is to solve ‘intuitive’ problems i.e. problems characterized by High dimensionality and no rules.  The above mechanism demonstrates a supervised learning algorithm based on a limited modelling of Neurons – but we need to understand more.

Deep learning allows computers to solve intuitive problems because:

• With Deep learning, Computers can learn from experience but also can understand the world in terms of a hierarchy of concepts – where each concept is defined in terms of simpler concepts.
• The hierarchy of concepts is built ‘bottom up’ without predefined rules by addressing the ‘representation problem’.

This is similar to the way a child learns ‘what a dog is’ i.e. by understanding the sub-components of a concept ex  the behavior(barking), shape of the head, the tail, the fur etc and then putting these concepts in one bigger idea i.e. the Dog itself.

The (knowledge) representation problem is a recurring theme in Computer Science.

Knowledge representation incorporates theories from psychology which look to understand how humans solve problems and represent knowledge.  The idea is that: if like humans, Computers were to gather knowledge from experience, it avoids the need for human operators to formally specify all of the knowledge that the computer needs to solve a problem.

For a computer, the choice of representation has an enormous effect on the performance of machine learning algorithms. For example, based on the sound pitch, it is possible to know if the speaker is a man, woman or child. However, for many applications, it is not easy to know what set of features represent the information accurately. For example, to detect pictures of cars in images, a wheel may be circular in shape – but actual pictures of wheels may have variants (spokes, metal parts etc). So, the idea of representation learning is to find both the mapping and the representation.

If we can find representations and their mappings automatically (i.e. without human intervention), we have a flexible design to solve intuitive problems.   We can adapt to new tasks and we can even infer new insights without observation. For example, based on the pitch of the sound – we can infer an accent and hence a nationality. The mechanism is self learning. Deep learning applications are best suited for situations which involve large amounts of data and complex relationships between different parameters. Training a Neural network involves repeatedly showing it that: “Given an input, this is the correct output”. If this is done enough times, a sufficiently trained network will mimic the function you are simulating. It will also ignore inputs that are irrelevant to the solution. Conversely, it will fail to converge on a solution if you leave out critical inputs. This model can be applied to many scenarios as we see below in a simplified example.

An example of learning through layers

Deep learning involves learning through layers which allows a computer to build a hierarchy of complex concepts out of simpler concepts. This approach works for subjective and intuitive problems which are difficult to articulate.

Consider image data. Computers cannot understand the meaning of a collection of pixels. Mappings from a collection of pixels to a complex Object are complicated.

With deep learning, the problem is broken down into a series of hierarchical mappings – with each mapping described by a specific layer.

The input (representing the variables we actually observe) is presented at the visible layer. Then a series of hidden layers extracts increasingly abstract features from the input with each layer concerned with a specific mapping. However, note that this process is not pre defined i.e. we do not specify what the layers select

For example: From the pixels, the first hidden layer identifies the edges

From the edges, the second hidden layer identifies the corners and contours

From the corners and contours, the third hidden layer identifies the parts of objects

Finally, from the parts of objects, the fourth hidden layer identifies whole objects

Image and example source: Yoshua Bengio book – Deep Learning

Implications for IoT

To recap:

• Deep learning algorithms apply to many areas including Computer Vision, Image recognition, pattern recognition, speech recognition, behaviour recognition etc
• Deep learning systems are possible to implement now because of three reasons: High CPU power, Better Algorithms and the availability of more data. Over the next few years, these factors will lead to more applications of Deep learning systems.
• Deep learning applications are best suited for situations which involve large amounts of data and complex relationships between different parameters.
• Solving intuitive problems: Training a Neural network involves repeatedly showing it that: “Given an input, this is the correct output”. If this is done enough times, a sufficiently trained network will mimic the function you are simulating. It will also ignore inputs that are irrelevant to the solution. Conversely, it will fail to converge on a solution if you leave out critical inputs. This model can be applied to many scenarios

In addition, we have limitations in the technology. For instance, we have a long way to go before a Deep learning system can figure out that you are sad because your cat died(although it seems Cognitoys based on IBM watson is heading in that direction). The current focus is more on identifying photos, guessing the age from photos(based on Microsoft’s project Oxford API)

And we have indeed a way to go as Andrew Ng reminds us to think of Artificial Intelligence as building a rocket ship

“I think AI is akin to building a rocket ship. You need a huge engine and a lot of fuel. If you have a large engine and a tiny amount of fuel, you won’t make it to orbit. If you have a tiny engine and a ton of fuel, you can’t even lift off. To build a rocket you need a huge engine and a lot of fuel. The analogy to deep learning [one of the key processes in creating artificial intelligence] is that the rocket engine is the deep learning models and the fuel is the huge amounts of data we can feed to these algorithms.”

Today, we are still limited by technology from achieving scale. Google’s neural network that identified cats had 16,000 nodes. In contrast, a human brain has an estimated 100 billion neurons!

There are some scenarios where Back propagation neural networks are suited

• A large amount of input/output data is available, but you’re not sure how to relate it to the output. Thus, we have a larger number of “Given an input, this is the correct output” type scenarios which can be used to train the network because it is easy to create a number of examples of correct behaviour.
• The problem appears to have overwhelming complexity. The complexity arises from Low rules base and a high dimensionality and from data which is not easy to represent.  However, there is clearly a solution.
• The solution to the problem may change over time, within the bounds of the given input and output parameters (i.e., today 2+2=4, but in the future we may find that 2+2=3.8) and Outputs can be “fuzzy”, or non-numeric.
• Domain expertise is not strictly needed because the output can be purely derived from inputs: This is controversial because it is not always possible to model an output based on the input alone. However, consider the example of stock market prediction. In theory, given enough cases of inputs and outputs for a stock value, you could create a model which would predict unknown scenarios if it was trained adequately using deep learning techniques.
• Inference:  We need to infer new insights without observation. For example, based on the pitch of the sound – we can infer an accent and hence a nationality

Given an IoT domain, we could consider the top-level questions:

• What existing applications can be complemented by Deep learning techniques by adding an intuitive component? (ex in smart cities)
• What metrics are being measured and predicted? And how could we add an intuitive component to the metric?
• What applications exist in Computer Vision, Image recognition, pattern recognition, speech recognition, behaviour recognition etc which also apply to IoT

Now, extending more deeply into the research domain, here are some areas of interest that I am following.

Complementing Deep Learning algorithms with IoT datasets

In essence, these techniques/strategies complement Deep learning algorithms with IoT datasets.

1)      Deep learning algorithms and Time series data : Time series data (coming from sensors) can be thought of as a 1D grid taking samples at regular time intervals, and image data can be thought of as a 2D grid of pixels. This allows us to model Time series data with Deep learning algorithms (most sensor / IoT data is time series).  It is relatively less common to explore Deep learning and Time series – but there are some instances of this approach already (Deep Learning for Time Series Modelling to predict energy loads using only time and temp data  )

2)      Multiple modalities: multimodality in deep learning. Multimodality in deep learning algorithms is being explored  In particular, cross modality feature learning, where better features for one modality (e.g., video) can be learned if multiple modalities (e.g., audio and video) are present at feature learning time

3)      Temporal patterns in Deep learning: In their recent paper, Ph.D. student Huan-Kai Peng and Professor Radu Marculescu, from Carnegie Mellon University’s Department of Electrical and Computer Engineering, propose a new way to identify the intrinsic dynamics of interaction patterns at multiple time scales. Their method involves building a deep-learning model that consists of multiple levels; each level captures the relevant patterns of a specific temporal scale. The newly proposed model can be also used to explain the possible ways in which short-term patterns relate to the long-term patterns. For example, it becomes possible to describe how a long-term pattern in Twitter can be sustained and enhanced by a sequence of short-term patterns, including characteristics like popularity, stickiness, contagiousness, and interactivity. The paper can be downloaded HERE

Implications for Smart cities

I see Smart cities as an application domain for Internet of Things. Many definitions exist for Smart cities/future cities. From our perspective, Smart cities refer to the use of digital technologies to enhance performance and wellbeing, to reduce costs and resource consumption, and to engage more effectively and actively with its citizens (adapted from Wikipedia). Key ‘smart’ sectors include transport, energy, health care, water and waste. A more comprehensive list of Smart City/IoT application areas are: Intelligent transport systems – Automatic vehicle , Medical and Healthcare, Environment , Waste management , Air quality , Water quality, Accident and  Emergency services, Energy including renewable, Intelligent transport systems  including autonomous vehicles. In all these areas we could find applications to which we could add an intuitive component based on the ideas above.

Typical domains will include Computer Vision, Image recognition, pattern recognition, speech recognition, behaviour recognition. Of special interest are new areas such as the Self driving cars – ex theLutz pod and even larger vehicles such as self driving trucks

Conclusions

Deep learning involves learning through layers which allows a computer to build a hierarchy of complex concepts out of simpler concepts. Deep learning is used to address intuitive applications with high dimensionality.  It is an emerging field and over the next few years, due to advances in technology, we are likely to see many more applications in the Deep learning space. I am specifically interested in how IoT datasets can be used to complement deep learning algorithms. This is an emerging area with some examples shown above. I believe that it will have widespread applications, many of which we have not fully explored(as in the Smart city examples)

I see this article as part of an evolving theme. Future updates will explore how Deep learning algorithms could apply to IoT and Smart city domains. Also, I am interested in complementing Deep learning algorithms using IoT datasets.

I elaborate these ideas in the Data Science for Internet of Things program  (modelled on the course I teach at Oxford University and UPM – Madrid). I will also present these ideas at the International conference on City Sciences at Tongji University in Shanghai  and the Data Science for IoT workshop at the Iotworld event in San Francisco

Please connect with me if you want to stay in touch on linkedin and for future updates

Follow us @IoTCtrl | Join our Community

Guest blog post by ajit jaokar

Often, Data Science for IoT differs from conventional data science due to the presence of hardware.

Hardware could be involved in integration with the Cloud or Processing at the Edge (which Cisco and others have called Fog Computing).

Alternately, we see entirely new classes of hardware specifically involved in Data Science for IoT(such as synapse chip for Deep learning)

Hardware will increasingly play an important role in Data Science for IoT.

A good example is from a company called Cognimem which natively implements classifiers(unfortunately, the company does not seem to be active any more as per their twitter feed)

In IoT, speed and real time response play a key role. Often it makes sense to process the data closer to the sensor.

This allows for a limited / summarized data set to be sent to the server if needed and also allows for localized decision making.  This architecture leads to a flow of information out from the Cloud and the storage of information at nodes which may not reside in the physical premises of the Cloud.

In this post, I try to explore the various hardware touchpoints for Data analytics and IoT to work together.

Cloud integration: Making decisions at the Edge

Intel Wind River edge management system certified to work with the Intel stack  and includes capabilities such as data capture, rules-based data analysis and response, configuration, file transfer and  Remote device management

Integration of Google analytics into Lantronix hardware –  allows sensors to send real-time data to any node on the Internet or to a cloud based application.

Microchip integration with Amazon Web services  uses an  embedded application with the Amazon Elastic Compute Cloud (EC2) service. Based on  Wi-Fi Client Module Development Kit . Languages like Python or Ruby can be used for development

Integration of Freescale and Oracle which consolidates data collected from multiple appliances from multiple Internet of things service providers.

Libraries

Libraries are another avenue for analytics engines to be integrated into products – often at the point of creation of the device. Xively cloud services is an example of this strategy through xively libraries

APIs

In contrast, keen.io provides APIs for IoT devices to create their own analytics engines ex (smartwatch Pebble’s using of keen.io)  without locking equipment providers into a particular data architecture.

Specialized hardware

We see increasing deployment  of specialized hardware for analytics. Ex egburt from Camgian which uses sensor fusion technolgies for IoT.

In the Deep learning space, GPUs are widely used and more specialized hardware emerges such asIBM’s synapse chip. But more interesting hardware platforms are emerging such as Nervana Systemswhich creates hardware specifically for Neural networks.

Ubuntu Core and IFTTT spark

Two more initiatives on my radar deserve a space in themselves – even when neither of them have currently an analytics engine:  Ubuntu Core – Docker containers+lightweight Linux distribution as an IoT OS and IFTTT spark initiatives

Comments welcome

This post is leading to vision for Data Science for IoT course/certification. Please sign up on the link if you wish to know more when launched in Feb.

Image source: cognimem

Follow us @IoTCtrl | Join our Community

Guest blog post by vozag
 

Emergence of IoT presents security challenges more challenging than any industrial systems have seen.

Open Web Application Security Project (OWASP) is a reputed international organization which focuses on improving the security of the software. It sponsors the hugely  popular Top ten project which publishes the top ten security risks for web applications all over the world.

The “OWASP Internet of Things (IoT) Top 10” project defines the top ten security surface areas presented by IoT systems. The project aims to provide practical security recommendations for builders, breakers, and users of IoT systems.

Last year HP which started this project used it as a baseline to evaluate top ten IoT devices which are were widely used and released a report. The study concluded that on an average each device studied had 25 vulnerabilities listed as a part of project.

The top 10 vulnerabilities impact of each vulnerability and the link in the order listed in project are given below:

Insecure Web Interface

Insecure web interfaces can result in data loss or corruption, lack of accountability, or denial of access and can lead to complete device takeover.

Insufficient Authentication/Authorization

Insufficient authentication/authorization can result in data loss or corruption, lack of accountability, or denial of access and can lead to complete compromise of the device and/or user accounts.

Insecure Network Services

Insecure network services can result in data loss or corruption, denial of service or facilitation of attacks on other devices.

Lack of Transport Encryption

Lack of transport encryption can result in data loss and depending on the data exposed, could lead to complete compromise of the device or user accounts.

Privacy concerns

Collection of personal data along with a lack of protection of that data can lead to compromise of a user's personal data.

Insecure Cloud Interface

An insecure cloud interface could lead to compromise of user data and control over the device.

Insecure Mobile Interface

An insecure mobile interface could lead to compromise of user data and control over the device.

Insufficient Security Configurability

Insufficient security configurability could lead to compromise of the device whether intentional or accidental and/or data loss.

Insecure_Software/Firmware

Insecure software/firmware could lead to compromise of user data, control over the device and attacks against other devices.

Poor Physical Security

Insufficient physical security could lead to compromise of the device itself and any data stored on that device.

Given all the buzz happening in the market around IoT, We looked at related projects in the crowd funding website Kickstarter.com to see how are IoT projects doing with respect to all the other ones.

We chose projects which have either “IoT” or “Internet of Things” either in their title or description and here are our findings.

The success rate of projects at Kickstarter is around 37.5%, for Technology projects it is 21% which is a lot less than the average success rate of projects. In Spite of this our analysis shows that the success rate of IoT projects is 44%, which is pretty good news. People are realizing the importance of IoT and are willing to fund the related projects.

The projects locations are almost concentrated in US and Europe with a few scattered in Asia and Australia

Because the projects are spread all over the world the goals of money to be raised were also in different currencies so to be able to analyse the monetary part we normalized all the numbers to US dollars.

The total sought out money for all the IoT related projects ( ongoing, successful and failed ) in Kickstarter is around $4.7 million and the actual pledged amount for the projects is around $1.5 million.

If you only consider the projects which have made it the total sought out and pledged amount is approximately $1.2 million. So only 2% of the pledged amount went to the unsuccessful projects which is usually the case with most of the projects on Kickstarter.

The average requested funding for all projects is around $60 thousand while the average funding requested by the successfully funded projects is around $44 thousand. For the failed projects it is $3500.

The top 10 successfully funded projects along with their links are given below

At the CES 2015, I was fascinated by all sorts of possible applications of IoT – socks with sensors, mattresses with sensors, smart watches, smart everything – it seems like a scene in sci-fi movies has just come true. People are eager to learn more about what’s happening around them and now they can.

While I was at there I attended a talk given by David Pogue – he is awesome. He pointed out that the prevalence of smartphone is the key to the realization of the phenomenon called “Quantified Self.” I agreed with him. Smart phones play a vital role as a hub where all our personal data converge and present, seamlessly. The fact that you carry your smartphone around all the time and that the screen size perfectly reveals all the information results in a catalyst for wearable devices, IoT or what we like to call it, Intelligence of Things.

It’s all relevant; Big Data, IoT, Wearable, Cloud Computing… While most data is uploaded to the cloud, the client devices are generally powerful enough that the computing can be decentralized. That said, small data (client side) and big data (server side) form an eco-system where small data triggers the knowledge base cultivated by big data and does the predictive analysis and decision making in a timely manner. Furthermore, your smartphone gathers versatile data and is able to analyze cross-app data to personalize your application settings. For example, what about optimizing navigation based on my physical condition? Or how about suggesting the best route according to my health along with the weather? These individual data records might be small, but collectively they enrich the content of analysis and contribute some amazing value. We at BigObject really appreciate this context of Big Data.

Marc Andreessen once said, “I think we are all underestimating the impact of aggregated big data across many domains of human behavior, surfaced by smartphone apps.” For us here at BigObject, the next big thing in big data is to find out a methodology that can link multiple data sources together and identify the meaningful connections between that data. Most importantly it must be responsive enough to deliver actionable insight and simple enough for people to adopt. That is the key to fulfill a connected world. 

Originally posted on Data Science Central

The Internet of Things (IOT) will soon produce a massive volume and variety of data at unprecedented velocity. If "Big Data" is the product of the IOT, "Data Science" is it's soul.

.

Let's define our terms:

.

Originally posted on AnalyticBridge

As big data becomes more of cliche with every passing day, do you feel Internet of Things is the next marketing buzzword to grapple our lives.

So what exactly is Internet of Thing (IoT) and why are we going to hear more about it in the coming days.

Internet of thing (IoT) today denotes advanced connectivity of devices,systems and services that goes beyond machine to machine communications and covers a wide variety of domains and applications specifically in the manufacturing and power, oil and gas utilities.

An application in IoT can be an automobile that has built in sensors to alert the driver when the tyre pressure is low. Built-in sensors on equipment's present in the power plant which transmit real time data and thereby enable to better transmission planning,load balancing. In oil and gas industry, it can help in planning better drilling, track cracks in gas pipelines.

IoT will lead to better predictive maintenance in the manufacturing and utilities and this is will in turn lead to better control, track, monitor or back-up of the process. Even a small percentage improvement in machine performance can significantly benefit the company bottom line.

IoT in some ways is to going to make our machines more brilliant and reactive.

According to GE, 150 Billion dollars in waste across major industries can be eliminated by IoT.

There can be questions that how is IoT different from a SCADA (supervisory control and data acquistion) systems which gets extensively used in the manfucturing industries.

IoT can be considered to be an evolution on the data acquisition part of the SCADA systems.

SCADA has been basically considered to be systems in silos with the data accessible to few people and not leading to long term benefit.

IoT starts with embedding advanced sensors in machines and collecting the data for advanced analytics.

As we start receiving data from the sensors , one important aspect that needs all the focus is the data transmitted correct or erroneous.

How do we validate the data quality.

We are dealing with uncertainty out here.

One of the most commonly used methods for modelling uncertainty is Bayesian networks.

Bayesian network is a probabilistic graphical model that represents a set of random variables and their conditional dependencies via a directed acyclic graph.

Bayesian networks can be used extensively in Internet of things projects to ascertain data transmitted by the sensors.

The Internet of Things may be giving over to the Internet of Everything as more and more uses are dreamed up for the new wave of Smart Cities.

In the Internet of Things, objects have their own IP address, meaning that sensors connected to the web can send data to the cloud on just about anything: how much traffic is rolling through a stoplight, how much water you’re using, or how full a trash dumpster is.

Cities are discovering how they can use these new technologies — and the data they generate — to be more efficient and cost effective in many different ways. And it’s a good thing, too; some estimates suggest that 66 percent of the world’s population will live in urban areas by the year 2050.

These are cutting edge ideas, but here are some of the most fascinating ways Smart Cities are using big data and the Internet of Things to improve quality of life for their residents:

• The city of Long Beach, California is using smart water meters to detect illegal watering in real time and have been used to help some homeowners cut their water usage by as much as 80 percent. That’s vital when the state is going through its worst drought in recorded history and the governor has enacted the first-ever state-wide water restrictions.
• Los Angeles uses data from magnetic road sensors and traffic cameras to control traffic lights and thus the flow (or congestion) of traffic around the city. The computerized system controls 4,500 traffic signals around the city and has reduced traffic congestion by an estimated 16 percent.  
• Xcel Energy initiated one of the first ever tests of a “smart grid” in Boulder, Colorado, installing smart meters on customers’ homes that would allow them to log into a website and see their energy usage in real time. The smart grid would also theoretically allow power companies to predict usage in order to plan for future infrastructure needs and prevent brown out scenarios.
• A tech startup called Veniam is testing a new way to create mobile wi-fi hotspots all over the city in Porto, Portugal. More than 600 city buses and taxis have been equipped with wifi transmitters, creating the largest free wi-fi hotspot in the world. Veniam sells the routers and service to the city, which in turn provides the wi-fi free to citizens, like a public utility. In exchange, the city gets an enormous amount of data — with the idea being that the data can be used to offset the cost of the wi-fi in other areas. For example, in Porto, sensors tell the city’s waste management department when dumpsters are full, so they don’t waste time, man hours, or fuel emptying containers that are only partly full.
• New York City is creating the world’s first “quantified community” where nearly everything about the environment and residents will be tracked. The community will be able to monitor pedestrian traffic flow, how much of the solid waste collected is recyclable or food waste, and air quality. The project will even collect data on residents’ health and activity levels through an opt-in mobile app.
• Songdo, South Korea has been conceived and built as the ultimate Smart City — a city of the future. Trash collection in the city is completely automated, through pipes connected to every building. The solid waste is sorted then recycled, buried, or burned for fuel. The city is partnering with Cisco to test other technologies, including home appliances and utilities controlled by your smartphone, and even a tracking system for children (using microchips implanted in bracelets).

This is just the beginning of the integration of big data and the Internet of Things into daily life, but it is by no means the end. As our cities get smarter and begin collecting and sending more and more data, new uses will emerge that may revolutionize the way we live in urban areas.

Of course, more technology can also mean more opportunities for hackers and terrorists. (Anyone see Die Hard 4, where terrorists hacked the traffic control systems in Washington, D.C.?) The threat that a hacker could shut down a city’s power grid, traffic system, or water supply is real — mostly because the technology is so new that cities and providers are not taking the necessary steps to protect themselves.

Still, it would seem that the benefits will outweigh the risks with these new data-driven technologies for cities, so long as the municipalities are paying attention to security and protecting their assets and their customers.

What’s your opinion? Are you for or against more integrated technologies in cities? I’d love to hear your thoughts in the comments below.

I hope you found this post interesting. I am always keen to hear your views on the topic and invite you to comment with any thoughts you might have.

About : Bernard Marr is a globally recognized expert in analytics and big data. He helps companies manage, measure, analyze and improve performance using data.

His new book is: Big Data: Using Smart Big Data, Analytics and Metrics To Make Better Decisions and Improve Performance You can read a free sample chapter here.

Originally posted on Data Science Central