GradOS

"Wireless data caps: Are usage based pricing schemes here to stay?" by By Larry Dignan for Between the Lines | March 10, 2009 -- 02:19 GMT (19:19 PDT)

Wireless data caps is a hot an fiery topic.  Usage based pricing schemes are here to stay: it's the American way.  It is no different than the way we consume electricity.  Communications services are important and we need them for many aspects of our lives.  We use wireless services for (a) communication, (b) entertainment, (c) work from home, etc.  But when compared to electricity is the pricing for wireless data fair?  I'd like to argue that the answer is no.  Data plans could be change to be a pay-for-use, but without minimums.  

The problem is that communications services, like gas or electricity, are supplied to us by private corporations with sustainable models dependent on profit-making.  Carriers are interested in making money and generating profit for their investors.  So its natural to charge what they can in order to increase their revenue.  Is that ethical or fair?  Perhaps no, but it is probably within the boundaries of the law.

Here to stay

These data caps are here to stay.  Just last year news outlets have published articles on this trend that is becoming the way of doing business.  

AT&T Ends All-You-Can-Eat

Verizon's and Comcast's data caps: Who wins and who loses? 

New Netflix iOS app capitulates to bandwidth caps

The other side of the coin

Title: uWave: Accelerometer-based personalized gesture recognition and its applications.

Author(s): Jiayang Liu, Lin Zhong, Jehan Wickramasuriya, Venu Vudevan

Journal: Pervasive and Mobile Computing; Pervasive and Mobile Computing, Volume 5, Issue 6, Pages 657-675

Year: 2009

DOI: 10.1016/j.pmcj.2009.07.007

Summary:

Automatic gesture recognition based on built-in accelerometers to be used in a wide range of devices, but especially on resource-constrained systems requiring minimal training.  Their work is intended to be used for personalized gesture recognition with applications in user authentication.  The contribution of this work is an improvement to current methods of gesture recognition that employ Hidden Markov Models based methods with 12 training samples.  In contrast, their scheme named uWave, requires a single training sample.   uWave also improves the quantization, helps suppress noise and reduces the computation load.

Fig. 1 and 2 typical examples of gestures for a hand-held device for remote control.

Fig. 3 Password gestures for critical authentication (two for each participant).

Major Results of the Study:

Gesture Recognition using the uWave algorithm reduces the training samples required for accurate and reliable personal gesture recognition of 98.6% with  template adaption.

uWave offers relatively fast recognition without requirement complicated optimization.

Materials and Methods:

The uWave scheme was implemented on multiple platforms that included smartphones, micro-controllers, and the Nintendo Wii.  The Nintendo Wii remote is connected wirelessly to a PC via Bluetooth.   Gesture recognition uses a library of 4480 stored gestures for eight gesture paters from eight participants over multiple weeks.  The smartphone implementation is prototyped on a Windows Mobile and iPhone.  The key is their uWave algorithm that uses template adaptation and template matching with dynamic time warping, a technique used in connected word recognition.

Fig. 4. uWave is based on acceleration quantization, template matching with DTW, and template adaptation.

What the Authors Learned:

uWave shows that personalized gesture recognition can achieve a success rate or accuracy of 98.6% with template adaptation.  Template adaptation is a technique used in speech recognition to handle temporal variations in gesture samples.

User Authentication gesture recognition still faces significant challenges, the results show that for non-critical applications the success rate is acceptable, but for critical applications, the visual disclosure is an issue.

Strong Points:

The evaluation of their algorithm in user authentication used an experimental group of size n=25 over a month.

Suggestion(s):

When presenting information that claims to be better than other, more established, methods the authors should show hard numbers indicating how the two compare.  Specifically, what the energy profile is on  a smartphone when the only sensor used is the accelerometer.  It would be beneficial to see the contrast between uWave computing profile to that of other systems, such as the visual method of recognition or one that more closely parallels this accelerometer-based method.

References for Further Exploration:

Emiliano Miluzzo , Tianyu Wang , Andrew T. Campbell, EyePhone: activating mobile phones with your eyes, Proceedings of the second ACM SIGCOMM workshop on Networking, systems, and applications on mobile handhelds, August 30-30, 2010, New Delhi, India

I have been following the work out of Prof. Campbell's lab, so I thought this would be an interesting paper to look at next, especially since it has a bold title.

Reference:

The Stanford Education Experiment Could Change Higher Learning Forever

Original Source:http://www.wired.com/wiredscience/2012/03/ff_aiclass//

Steven Leckart, Wired Science

Wired.com, March 20, 2012

  When I read the article, down somewhere in middle of it the following images and concepts began to take shape in my head: medieval and early history's masters and apprentices, master Shifu, etc.  The reason this came to mind is because I see both individual teachers (the best in their fields) and clusters of both teachers and students centered around a topics they are truly interested in and enjoy discussing at length, being as reachable as masters were to their apprentices.  Individuals interested in getting the best training or information anywhere in the world with internet access, should be able to do so.

I see on-line training as a disruptive trend, because I too envision universities taking new shape in decades to come.  The new shape I see or envision is one where people, in their formative years, are taught to develop an interest in a field (or fields) and then go out and seek training from the masters.  These new masters will be more like Thrun, in that they will be the best teachers not because of their university affiliation, but either because of who they are.  Who they are is defined by their accomplishments, their talent, and personal traits (great communicators, patience to teach, etc.).  I also envision clusters of people being sources for learning.  The clusters or groups are the equivalent of forums that help us learn and master new skills.  The brick and mortar builds representing this and that university today will still be there, but so much as the university that we attend to earn our degrees from, but the physical places were we learn from teachers and masters that are either there or any where in the world.  

It will take experimentation: to show that we have learned and prepared ourselves sufficiently from on-line learning, it will involve taking bold steps forward.  We need to show that an individual trained by the worlds best, is as good as those graduating from the most expensive and and best universities of today.    We need to break up the Yale, Stanford, Harvard, and MIT's of the world into pieces in order to free their best to teach not just an elite group of people (as they do today), but anyone who wants to learn from them.

I want a world, where if I want to learn AI I can go learn it from Sebastian Thrun and if I want to learn quantum gravity, then I can learn that from Brian Greene.  I am looking forward to a disruption in higher education where a new society values what you've learned, how well you've learnt it, from whom you've learned it from, rather than what university you attended.

University Education for Those Who Live in the Jungle

The University of the People is an other example of an online university that offers degrees in CS and is not only online, but it is free.  This university is an example of where anyone in the world can be taught by the best in their field.  Some of the online teachers are retired professors.  Some of which might be exceptional teachers even into their retirement years.   According a report on NPR, the skeptics point out that a model like this one might not be sustainable.  Another problem pointed out, rather significant, is how do we trust that the person taking an exam (out there in the jungle) is in fact the right person him or herself?  

Caution

In addition to the problem listed above, there are many other problems with an online model.  In this country alone, recently there's been many scandals where online universities abuse the system around financial aid fraud.  Which some view as the problem of government enticing some to take out loans, because it is easier to be admitted to some online schools.

CS For All

Title: A Closed-Loop Approach for Improving the Wellness of Low-Income Elders at Home Using Game Consoles

Authors:  Jong Hyun Lim, Andong Zhan, JeongGil Ko, Andreas Terzis, Sarah Szanton, and Laura Gitlin,

Affiliation:  Johns Hopkins University, MD, USA

Presented/Published:  IEEE Communications Magazine, Jan. 2012 

Digital Object Identifier (DOI)::  10.1109/MCOM.2012.6122531

Summary:

HealthOS, a middleware platform that sits on the cloud, is a new approach that aims to minimize the need for user-intervention by caregivers (i.e. nurses or family relatives), and improves self-management that yields more participation by unattended elderly (typically living at home and especially for those with low-income) in performing physical performance tests design to reduce the risk of falls and detect the chances of getting a heart attack. HealthOS is also a development platform.  It allows integration of multiple healthcare monitoring devices (or fitness monitors) and facilitates easy access to the data for analysis, visualization, presentation, and integration with existing healthcare record systems.

Statement of the Research Problem: 

The number of elders with complex chronic conditions is rising as the number of middle-aged adults reporting difficulty with physical function increases and as more elders with disabilities prefer to live at home; these factors combined are two of the major contributors to rising healthcare costs. Also, these two factors can lead to poor quality of life (especially for those with low income).

After developing a system that combines balance exercises (to prevent fall risk) with monitors that help assess the risk of heart-attacks, coupled with periodic nurse visits, and conducting a set of ’Short Physical Performance Battery of tests for the participants the results showed low participation (some only practiced when nurses visited their home). So, the key question is: ”how do we close the intervention loop so that we can minimize the caregivers’ workload, include family members, and increase the quality of care at the same time?” 

Description of the Research Procedures:

Their stated aim is: ”to improve the process of various applications in the healthcare domain where user- interventions are necessary.” Their approach uses a gaming system (e.g. Nintendo Wii) with a physical exercise app designed to motivate, offer various levels of difficulty, have an UI suited for the elderly, offer voice interaction, and various ways of showing the progress or results. A tool to interact with caregivers, a network of similar patients, and family members is also a requirement so use of a smartphone is an important element in their proposed solution.

To combine the weight measures from the gaming system with data from heart rate monitors and see the correlation for detecting the chances of a heart-attack, for example, they have developed HealthOS. HealthOS is a middleware running on the cloud. It collects data from pervasive healthcare devices, inte- grates healthcare applications, and provides a development platform through RESTful APIs.

In a nutshell, HealthOS has four layers of function: 1) device drivers 2) a database layer, 3) a pipeline layer (data routing/packaging) 4) management, security (privacy), and mobility. At the edges of this system on one side are: the health monitoring devices, and at the other are data visualization, data presentation, and healthcare record repositories interfaces. They describe a simple and easy way to create drivers and pipelines, but only show an example where the Nintendo Wii is used. The description of encryption (for data access) doesn’t compare their approach to others nor wether it addresses HIPAA compliance, but they do cite an article that discusses Ciphertext-Policy Attribute based Encryption (shown to have been used in HIPAA compliance).

The article does not present how this tool or approach actual performs. There is no research that was actually carried out, thus no results are presented in the article.

Data Analysis:

There are no results from actual test subjects presented in this article.

Conclusions:

Their hypothesis is never tested. They designed an innovative tool, but in this article it is not put into action. Looking at the first author’s website, there are two submitted articles for publication listed. Which might suggest that actual results and data analysis are presented.

Pros:

A middleware platform that enables multi-device sensing sounds like a great idea 

The authors briefly describe an easy and simple way to create drivers for new devices and pipelines for these

HealthOS employs encryption

Excellent mobile tool DailyAlert (Android app) that connects to a cloud and offers Google Cloud to Device Messaging. It is a mobile intervention system design to provide alert notifications to both patients and caregivers and to a patient’s support network. 

Cons:

The article does not present evidence on how falling or heart attacks contribute to rising healthcare costs

They don’t show or cite previous work showing that short physical perfor- mance exercises that contributes to better balance, reduces falls, or reduces the incidents of heart attacks

There is no evidence in the article on how their approach (HealthOS) minimizes the need for user-intervention by caregivers and improves self- management that yields more participation by unattended elderly in per- forming physical performance tests.

Designing something like this for elderly with low-income is too broad, in other words many elderly with low-income can be further classified into those that are 1) internet savvy, with access to gaming gear, able to afford a smartphone, 2) those on the other side of the digital divide, who would not be able to afford a smartphone nor gaming equipment simply because of factors like race, socioeconomic status, etc., and last, but not least, 3) elderly with complicated medical conditions (or not) often have or perceive barriers to caring for themselves that can be far more complicated than one can imagine. 

The items listed in the Cons section are areas for improvement.

Related Papers:

I visited the first-author's website an did notice that he has some papers out for publishing.  I am looking forward to seeing how that work augments or complements this paper.

Darko Kirovski and Nuria Oliver and Mike Sinclair and Desney O. Tan, "Health-OS:  a position paper," International Conference on Mobile Systems, Applications, and Services," 2007, pp 76--78, doi: 10.1145/1248054.1248077

Matt Might is an Assistant Professor at the University of Utah and on his blog he has written an article on the general topic of undergraduate computer science: "What every CS major should know."   It is an article in which he tries to answer the question by providing a perspective framed around the conjunction of four concerns, none of which I agree with, but one has to start somewhere: what should every student know to 1) get a good job, 2) maintain lifelong employment 3) enter graduate school, and 4) benefit society?  But, in listing what every CS major should know as they finish their undergraduate program I do agree with many of the items (skills) he lists.

Touch the hardware

There is one thing (skill) I did not see listed, that every CS major should be familiar not only with architecture and an understanding of the caches, buses, and memory management, from a bookish point of view, but also be comfortable touching hardware.  This applies to CS majors going into product development (electronics hardware development, i.e. mobile phone design, etc).  By hardware I mean: development boards, or the hardware on which their software will run.  I've experienced many software developers (many of which were CS majors) that were not comfortable "touching" the hardware.  The problem is that a significant amount time is lost when software developers do not want to touch hardware and toss development hardware over to the EE side of the fence.  Being able to troubleshoot hardware by using an oscilloscope (or similar instrument) can save a development team time headaches and animosity between HW and SW developers. 

Should CS majors learn to work more intimately with hardware during their undergraduate training?  Perhaps yes, but I think that it is also the responsibility of employers to provide this kind of training.  Leaving this to an employer (hardware manufacturer) is a better idea, because the training can be tailored to the products they design and manufacture.

On Parallelism