Isaac Kofi Owusu1, Fred Adomako-Boateng2, Fred Kueffer3, Molly Guy3, Chemuttaai Lang’at3, Daniel B Grossman4, Teri Whitman3, Keith Holloman3, Douglas A Hettrick3* and Opare-Sem O1 1Department of Medicine, School of Medical Sciences, College of Health Sciences, Kwame Nkrumah University of Science and Technology, Kumasi, Ghana 2Regional Health Directorate, Ghana Health Service, Kumasi, Ghana 3Medtronic, Inc, Minneapolis, MN, USA 4Mayo Clinic, Rochester, MN, USA

Background: Logistic and socioeconomic barriers limit effective blood pressure (BP) control in many parts of Sub-Saharan Africa, including the Republic of Ghana. We tested a novel hypertension management model of care designed for resource-limited settings. Methods and Results: The “Akoma Pa” model was developed using human-centered design methodology involving patients, physicians, and nurses. The model consisted of a mobile tablet, BP machine and a novel software application in a unique platform to allow for longitudinal patient management. Patients were provided with a tailored hypertension management plan based on their enrollment comorbidities and risk factors. A cohort of 150 hypertensive patients (57±8 years; 73% female) accessed regular blood pressure assessments at a local pharmacy and received real-time automated feedback based on their individualized plan. On the mobile application, clinicians were able to view patient data, provide patients with feedback via SMS on their condition, and write electronic prescriptions which could be accessed by participating pharmacies. Average baseline BP was 135±18/84±10 mmHg in the overall cohort and 153±13/90±11 mmHg in the subgroup with uncontrolled hypertension (n=58). After 6 months of voluntary weekly monitoring, systolic blood pressure decreased significantly (p<0.01) in the overall cohort (-4.7±18.7 mmHg) and in the uncontrolled subgroup (-15.2±17.6mmHg). Systolic blood pressure remained constant in the sub group with controlled pressure at baseline. The proportion of the population with uncontrolled hypertension decreased from 39% to 27% (p=0.01). Patient compliance with weekly BP assessments was 61% and 2,855 BP assessments were conducted. During 33 of the 2,855 BP assessments (1% of pharmacy visits), the software application directly referred patients to a health facility (33 visits in 25 patients). Improvement in overall health awareness was reported in 82% of the participants and 95% of participants indicated a desire to continue using this model in the future. Conclusions: Compliance and satisfaction with this multifaceted hypertension care model were high and led to significant and sustained decreases in blood pressure in this West African hypertensive population.

Correlations are poor between clinical pain and tissue abnormalities in many disorders, including rheumatoid arthritis (RA), and fibromyalgia syndrome (FM). Neither joint swelling, articular cartilage loss, nor inflammation predict well the symptom magnitude of RA and FM patients. Additionally, lack of easily identifiable tissue abnormalities in FM makes these patients' pains difficult to understand. However, recent findings of similar neuroplastic changes in RA and FM patients suggested common mechanisms relevant for clinical pain including peripheral and central sensitization. Thus we hypothesized that QST would predict clinical pain better than joint abnormalities in patients with RA or FM. 32RA and 69FM patients were evaluated in Rheumatology clinic. All patients received 20s pressure stimuli to the hands at 400kPa. During the stimuli continuous experimental pain ratings were obtained using an electronic visual-analogue-scale (VAS) to characterize peripheral pain sensitivity as well as 40s pain-aftersensations (AS). Area-under-the-curve (AUC) was used for statistical analyses. Tenderness and swelling of 74 peripheral joints was assessed. Average clinical pain of 30RA and 65FM patients was 1.94 (2.02) and 4.14 (2.38) VAS units, respectively. AUC of 20s mechanical stimuli + 40s AS was 297.3 and 1,086.2 units for RA and FM patients. RA patients reported 4.5 (5.7) tender and 4.7 (4.3) swollen joints at the upper and 1.3 (2.5) and 0.9 (1.3) joints at the lower extremities. The FM patients reported neither tender nor swollen joints. All group differences were statistically significant(p<.001). Clinical pain intensity of RA and FM patients significantly correlated with AUC (r=.45)(p<.01) but not with joint symptoms (p>.05). Specifically, the amount of joint tenderness and swelling did not predict clinical pain of RA patients. Thus, mechanical hyperalgesia seems to be important for both RA and FM pain. Future evaluations and treatments of RA and FM patients should focus on this important pain mechanism.

A new boson with mass ~125 GeV and properties similar to the Standard Model Higgs has been discovered by both the ATLAS and CMS collaborations, with significant observation in the ZZ ∗ → 4ℓ and γγ channels. In this work we ask whether the signals in these two channels can be due primarily to two distinct resonances, each contributing dominantly to one channel. We investigate this question in the framework of a 2HDM and several of its extensions. We conservatively find that such a scenario is not possible in a pure 2HDM, nor under the addition of vector-like quarks, but is allowed when adding one or two top-like scalars, if one allows for sub-one tan β. The resonances in the diboson and diphoton channels can then be two scalars, or a scalar and a pseudoscalar, respectively. In each viable case, we further find the expected future deviations in the diboson, diphoton, \( b\overline{b} \) and ττ rates, which will be useful in excluding the two-resonance scenario.

The new MEG bound on BR(mu -> e gamma) provides the strongest upper bound on the scale of gauge mediation of supersymmetry breaking. If, in the future, this decay is observed by MEG, the mediation scale will become known to within 1 order of magnitude, and the seesaw scale will be constrained. In such a case, contributions from Planck mediated supersymmetry breaking are likely to be non-negligible, and an interpretation in terms of purely seesaw parameters will be impossible. The recent evidence for vertical bar U-e3 vertical bar similar to 0.15 further sharpens the predictions of gauge-mediated supersymmetry breaking.

The past few years have seen a growing trend in application development toward "web applications", a fuzzy category of programs that currently (but not necessarily) run within web browsers, that rely heavily on network servers for data storage, and that are developed and deployed differently from traditional desktop applications. Where (typical) traditional applications are compiled pieces of code, written in arbitrary languages, that implement both an application's user interface and its functionality, web apps by contrast are written in three interpreted languages: HTML to define the structure or content of the UI, CSS to define the appearance, and JavaScript (JS) to define the behavior. These three languages feel nothing alike, and are used for different facets of the applications. The last decade has also seen the rise of Mozilla Firefox, a web browser whose UI and functionality are themselves written in (dialects of) HTML, CSS, and JS, making Firefox one of the first fully-fledged web apps. Part of Firefox's appeal is its strong support for extensions, which are downloadable, third-party pieces of code (i.e., not written by Mozilla or with Mozilla's cooperation) that enhance the browser with additional functionality or customizations. Firefox extensions are wildly popular: over six thousand distinct extensions have been downloaded over 2.5 billion times [193], and all major browser vendors have added varying degrees of support for extensions to their own products. Crucially, these extensions are also written in HTML, CSS, and JS: writing an extension feels fundamentally similar to writing a web page or web app. Thanks to the dynamic, interpreted nature of these three languages, it is mostly straightforward to incorporate the contents of an extension into the existing browser. There are, however, some caveats. Not all programs are equally amenable to post-hoc extension, and there are currently no guarantees that multiple extensions do not conflict, destabilizing each other or the base browser. In this dissertation, I aim to provide better support for rich extensibility for web apps. In particular, I claim that Language-specific extension mechanisms are needed for each of HTML, CSS, and JS, and such mechanisms are needed for building useful diagnostic tools to address inter-extension conflicts. To support this thesis, I first present C3, the "Cloud Computing Client", an implementation of the HTML/CSS/JS platform architected explicitly to support experimentation with extensibility. I then define two such extension mechanisms for HTML and for JS: overlays and aspects, respectively. I develop conflict analyses for HTML overlays, and evaluate them on a sample of Firefox extensions. Conflict analyses for JS are sketched, and extension mechanisms for CSS are left for future work.

Models of minimal lepton flavor violation where the seesaw scale is higher than the relevant flavor scale predict that all lepton flavor violation is proportional to the charged lepton Yukawa matrix. If extra vectorlike leptons are within the reach of the LHC, it will be possible to test the resulting predictions in ATLAS/CMS.

XML (the eXtensible Markup Language) is on the road to becoming a standard for computerized data presentation. With its ability to define a database’s structure within the data document itself (or within a small, attached Document Type Definition file), XML allows web visitors to download and view a site’s data with no previous knowledge of the data format. Thus, XML is a powerful language for rapidly obtaining and incorporating structured and semi-structured data from around the world. Several generations of data browsers, query languages, and query engines have already been deployed for XML, culminating most recently in the introduction of the Quilt query language [8]. Incorporating well-studied data storage and querying schemes from a variety of authors [4] [7] [5] [9] [6], we have implemented the first Quilt query engine (to our knowledge) — Quilt2Sql. At its heart is a relational storage schema and PostgreSQL relational query interface. In this paper we present the important components of our engine: an XML parser, XML-to-relational translator and storage schema, Quilt query parser, Quilt to SQL query translator, and XML result formatter. It should be noted that the Quilt language is rapidly changing: at the time of this publication, the Quilt grammar we employed was already obsolete.

"This book is essential medicine for the pandemic of scientific illiteracy. The architects of the explosive growth of science-based technology must communicate as never before and there is now a lucid guide." --Leon Lederman, 1988 Nobel Prize in Physics "This is an invaluable guide for scientists seeking to learn how to better communicate with--and through--the media. The book gives the kind of insight into new operations that will allow researchers to better understand the process of and to feel more comfortable talking with reporters. And the timing is just right--now more than ever we need an improved public understanding of science and the way it affects our lives." --Deborah Blum, 1992 Pulitzer-Prize winner for beat reporting on primate research. In this book, Richard Hayes and Daniel Grossman draw on their expertise in public relations and journalism to empower researchers in a variety of fields to spread their message on their own terms. The authors provide tips on how to translate abstract concepts into concrete metaphors, craft soundbites, and prepare for interviews. For those looking for a higher profile, the authors explain how to become a reporter's trusted source--the first card in the Rolodex--on controversial issues. A must-read for all scientists, this book shows how it IS possible for the discoveries that hibernate in lecture halls and academic journals to reach a broader audience in a way that is accurate and effective. Richard Hayes is media director of the Union of Concerned Scientists, an independent nonprofit alliance of more than 10,000 scientists and citizens. He lives in Arlington, Virginia. Daniel Grossman is a science journalist, radio and web producer, and educator who has covered the science beat for more than eighteen years. He lives in Watertown, Massachusetts.

Bayesian networks are a powerful probabilistic representation, and their use for classification has received considerable attention. However, they tend to perform poorly when learned in the standard way. This is attributable to a mismatch between the objective function used (likelihood or a function thereof) and the goal of classification (maximizing accuracy or conditional likelihood). Unfortunately, the computational cost of optimizing structure and parameters for conditional likelihood is prohibitive. In this paper we show that a simple approximation---choosing structures by maximizing conditional likelihood while setting parameters by maximum likelihood---yields good results. On a large suite of benchmark datasets, this approach produces better class probability estimates than naive Bayes, TAN, and generatively-trained Bayesian networks.

Research efforts into the engineering of real-time, distributed, virtual environments for use over the Internet have flourished since DARPA’s first steps in this direction fifteen years ago. The video game “Quake” (from idSoftware), having been the first successful game in this genre, has seen its interface and network usage characteristics become something of a standard in the network entertainment world. As Quake-like games which promise more detailed character skins, richer environment textures, and “realer” physics are released in rapid succession, such applications are quickly becoming the most data-intensive, latency-abhorring, and generally demanding distributed programs running on today’s networks. With this in mind, we have chosen the freely available “Quake I” and “QuakeWorld” sources as the subjects for a series of experiments to determine what transport and network layer strategies result in the best game play for the user, and why. In modifying these sources, we have attempted to define the term “acceptable gameplay” in networked first person shooters, and obtained quantitative data capturing this notion. Furthermore, we have quantitatively demonstrated that the later QuakeWorld code exhibits, as predicted, better network performance than the earlier Quake I code.

Two learning ensemble methods, Bagging and Boosting, have been applied to decision trees to improve classification accuracy over that of a single decision tree learner. We introduce Bagging and propose a variant of it — Improved Bagging — which, in general, outperforms the original bagging algorithm. We experiment on 22 datasets from the UCI repository, with emphasis on the ensemble’s accuracy as the main performance measure. The variant of Bagging that we propose utilizes the “out of bag” samples in determining a decision tree’s voting power, whereas in the classical Bagging algorithm, all trees have the same voting power. Our proposed algorithm creates bootstrap samples from the training data according to a forced probability distribution that emulates random sampling with replacement. We thus achieve a uniformly distributed training set for each base learner, and also identically-sized “out of bag” verification sets for each learner. Our 10-fold cross-validation results and single runs on large data sets show our novel Bagging variant to improve classification accuracy relative to both the original Bagging ensemble and standard Boosting.

Two learning ensemble methods, Bagging and Boosting, have been applied to decision trees to improve classication accuracy over that of a single decision tree learner. We introduce Bag- ging and propose a variant of it | Improved Bagging | which, in general, outperforms the original bagging algorithm. We experiment on 22 datasets from the UCI repository, with emphasis on the ensemble's accuracy as the main performance measure. The variant of Bag- ging that we propose utilizes the \out of bag" samples in determining a decision tree's voting power, whereas in the classical Bagging algorithm, all trees have the same voting power. Our proposed algorithm creates bootstrap samples from the training data according to a forced probability distribution that emulates random sampling with replacement. We thus achieve a uniformly distributed training set for each base learner, and also identically-sized \out of bag" verication sets for each learner. Our 10-fold cross-validation results and single runs on large data sets show our novel Bagging variant to improve classication accuracy relative to both the original Bagging ensemble and standard Boosting.