Adaptive Load Balancing for HPC Applications

Gopalakrishnan Menon, Harshitha Menon ProQuest Dissertations & Theses University of Illi 2016 해외박사(DDOD)

소속기관이 구독 중이 아닌 경우 오후 4시부터 익일 오전 9시까지 원문보기가 가능합니다.

One of the critical factors that affect the performance of many applications is load imbalance. Applications are increasingly becoming sophisticated and are using imbalance. Moreover, systems are becoming more complex. The number of cores per node is increasing substantially and nodes are becoming heterogeneous. High variability in the performance of the hardware components introduces further imbalance. Load imbalance leads to drop in system utilization and degrades the performance. To address the load imbalance problem, many HPC applications employ dynamic load balancing algorithms to redistribute the work and balance the load. Therefore, performing load balancing is necessary to achieve high performance. Different application characteristics warrant different load balancing strategies. We need a variety of high-quality, scalable load balancing algorithms to cater to different applications. However, using an appropriate load balancer is insufficient to achieve good performance because performing load balancing incurs a cost. Moreover, due to the dynamic nature of the application, it is hard to decide when to perform load balancing. Therefore, deciding when to load balance and which strategy to use for load balancing may not be possible a priori. With the ever increasing core counts on a node, there will be a vast amount of on-node parallelism. Due to the massive on-node parallelism, load imbalance occurring at the node level can be mitigated within the node instead of performing a global load balancing. However, having the application developer manage resources and handle dynamic imbalances is inefficient as well as is a burden on the programmer. The focus of this dissertation is on developing scalable and adaptive techniques for handling load imbalance. The dissertation presents different load balancing algorithms for handling inter and intra-node load imbalance. It also presents an introspective run-time system, which will monitor the application and system characteristics and make load balancing decisions automatically.

THE CULTURAL HISTORY OF M. MAYEUX IN NINETEENTH-CENTURY FRENCH ART AND LITERATURE. (VOLUMES I AND II) (JULY MONARCHY, SATIRE, CARICATURE)

MENON, ELIZABETH KOLBINGER University of Minnesota 1994 해외박사

Ashamed of our nakedness : Is there ever a naked body? Ambivalence in contemporary Indian expressive aesthetics

Menon, Rekha Telakat State University of New York at Binghamton 2001 해외박사

(The) evolution of South Korea's broadband convergence network, 2004-2007

Menon, Siddhartha Shankar UMI Dissertation Publishing 2010 국내박사

Impact of sgRNA sequence and secondary structure on CRISPR-Cas activity

Menon, A Vipin  한양대학교 대학원 2022 국내박사

Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR)-Cas 시스템은 포유류 체세포 유전학에서 게놈 공학에 혁명을 일으켰습니다. 상당한 연구들이 Spacer를 개선하거나 SpCas9 및 scaffold 서열(소위 tracrRNAs)을 조작함으로써 single guide RNA (sgRNA)-매개 편집의 효율을 향상시키는 것이 집중되었습니다. 초기 조사는 CRISPR-Cas 시스템의 표적 부위를 평가하기 위해 sgRNA의 spacer 서열을 사용하는 computational model의 개발로 이어졌고, 이로써 효율적인 guide RNA의 일반적인 설계 원리를 제시했습니다. 효율적인 sgRNA 설계를 위한 computational tool들은 일반화 문제를 나타내며 최적의 결과를 종합적으로 제공하지 않습니다. 또한, 이전 연구에서는 sgRNA 편집 효율 측면에서 2차 구조 및 scaffold 서열의 역할을 이해하기에는 한계가 있었습니다. 효율적인 sgRNA를 식별하는 설계 원리를 개선하기 위해서 4가지 다른 CRISPR 시스템에서 이를 확인하는 machine learning algorithm인 Elastic Net Logistic Regression (ENLOR)을 활용하는 Comprehensive Guide Designer (CGD)를 개발했습니다. CGD에는 편파적이지 않은 방식의 CRISPRi, CRISPRa, CRISPR-SpCas9 및 CRISPR-Cas12a (각각 CGDi, CGDa, CGD9 및 CGD12a)에서 생성한 공개 dataset로 훈련된 특정 model들이 포함되어 있습니다. 독립적인 test dataset로 평가했을 때 훈련된 CGD model 은 gRNA의 효능을 예측하는데 기존 방법을 능가했습니다. sgRNA spacer 서열의 Minimum Free Energy(MFE)를 보여주는 CGD model은 다양한 CRISPR-Cas 시스템에서 유전자 편집에 중요한 역할을 합니다. 연구 결과를 바탕으로 high throughput dataset에서 spacer와 scaffold 사이의 상호 작용을 분석하여 sgRNA 2차 구조의 역할을 평가했습니다. 이러한 분석을 통해서 대부분의 spacer-scaffold 상호작용은 편집 효율을 상당히 감소시키지만 소수의 상호작용은 고도로 활성화한 sgRNA의 효율을 향상시켰습니다. SpCas9 변이체와 base editor에 대한 추가 분석은 다양한 SpCas9 활성에 대한 sgRNA의 중요한 2차 구조를 기계적인 유사성과 차이점을 드러냈습니다. 연구 결과는 SpCas9, SpCas9 변이체 및 base editor 시스템에서 구조적 특징을 이용하여, 고도로 활성화한 sgRNA를 예측할 수 있는 CNN 기반 deep learning model을 통해 spacer-scaffold 상호작용을 정리합니다. 따라서 이 연구는 sgRNA 서열의 구조적 메커니즘에 대한 더 발달된 이해를 제공할 뿐만 아니라 CRISPR 시스템의 현재 치료 가치를 향상시킬 수 있는 고효율 sgRNA 설계하는 데 도움이 될 것입니다. The Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR)-Cas systems revolutionized genome engineering in mammalian somatic cells genetics. Significant research has focused on increasing the effectiveness of single guide RNA (sgRNA)-mediated editing by either refining the spacer or engineering SpCas9 and scaffold sequences (so-called tracrRNAs). Initial investigations led to the development of computational models using a spacer sequence of sgRNA to assess the target site of the CRISPR-Cas system, thereby laying out general design principles of efficient guide RNAs. The computational tools developed for designing efficient sgRNA exhibit generalization issues and do not comprehensively provide optimal results. Moreover, previous studies were limited in understanding the role of secondary structure and scaffold sequence in the sgRNA editing efficiency. To improve the design principle of identifying efficient sgRNAs, I developed Comprehensive Guide Designer (CGD), which utilizes the machine learning algorithm, Elastic Net Logistic Regression (ENLOR), that identifies efficient sgRNA across four different CRISPR systems. CGD contains specific models trained with public datasets generated by CRISPRi, CRISPRa, CRISPR-SpCas9, and CRISPR-Cas12a (designated as CGDi, CGDa, CGD9, and CGD12a, respectively) in an unbiased manner. When evaluated with independent test datasets, the trained CGD models outperformed the pre-existing methods in predicting the efficacy of gRNAs. The CGD model showcased MFE (Minimum Free Energy) of sgRNA spacer sequence, which plays a significant role in gene editing across different CRISPR-Cas systems. Based on the findings, I assessed the role of sgRNA secondary structure by analyzing the interactions between spacer and scaffold across high throughput datasets. My investigation showed that most spacer-to-scaffold interactions considerably reduce editing efficacies, but a handful of interactions enhanced the efficacies in highly active sgRNAs. Further analysis of SpCas9 variants and base editors revealed mechanistic similarity and dissimilarity of the significant secondary structures of sgRNAs over different SpCas9 activities. The findings incorporate spacer-scaffold interaction as structural features into a CNN-based deep learning model that could predict highly active sgRNAs in SpCas9, SpCas9 variants, and base editing system. My study thus will provide a better understanding of the structural mechanism of sgRNA sequences but also helps to design highly efficient sgRNAs which could enhance the present therapeutic value of the CRISPR system.

Determinants of parental beliefs: The role of proximal influences in the maintenance and revision of parental beliefs

Menon, Roshni The University of Arizona 2006 해외박사(DDOD)

소속기관이 구독 중이 아닌 경우 오후 4시부터 익일 오전 9시까지 원문보기가 가능합니다.

Culture has been recognized to play an important role in the formation of parental beliefs, but the question still remains of whether beliefs are maintained or revised over time, and how. The present study examined how proximal influences impacted parental beliefs in an immigrant sample of parents, the thesis being that distal influences have more to do with the formation of parental beliefs while proximal influences have more to do with maintaining or revising them. Effects of the proximal influences of education, occupation status, information networks, and parental agreement about childrearing, on parental beliefs of Mexican-origin fathers and mothers around cultural values of familism/respeto, simpatia, and individualism were tested longitudinally. The research questions were two-fold in nature, looking at within-time effects of the proximal influences on parental beliefs; as well as over-time effects of proximal influences on change in parental beliefs. The within-time questions were answered using hierarchical regression analyses while the over-time questions were answered using repeated measures MANCOVAs. Overall, the beliefs of parents in this study were seen to not change significantly over the course of the three years that they were assessed, and so the study did not yield the results expected in terms of the effects of proximal influences on parental beliefs. However, information networks and fathers' occupation status did emerge as promising proximal influences on parental beliefs, and the results also revealed maternal beliefs to be more responsive to the proximal influences of education, fathers' occupation status, information networks, and parental agreement about childrearing, than paternal beliefs.

Men on the Edge: Suicide, Masculinity, and Representation in Transnational Popular Culture

Menon, Rajiv Kannan New York University 2015 해외박사(DDOD)

소속기관이 구독 중이 아닌 경우 오후 4시부터 익일 오전 9시까지 원문보기가 가능합니다.

Focusing on the transnational relations produced through globalization and militarism among the United States, South Asia, and the Middle East, this dissertation explores how literal, potential, and symbolic acts of masculine suicide challenge dominant categories of race, class, and citizenship. In order to confront these concerns, this dissertation examines what I call the "identitarian suicide"-the suicide that is rationalized and explained through collective terms of identity, rather than an individualized pathology. Within popular discourses, for example, the visibility of groups like poor, rural farmers in India or Muslim American men is increasingly defined by the perceived risk of suicide among these communities, thus entailing that suicidal potential is central to cultural construction of these forms of identity. However, rather than functioning as an inherent characteristic of certain identity categories, I suggest suicide signals the failures of representation afforded by these classifications, and accordingly points to unheard forms of self-depiction that can potentially challenge dominant structures of power. My particular interest in masculine suicide arises from the ways that globalization and militarism appear as processes enacted in active, male terms, which consequently undergird the need to obscure the subversive potential of non-normative masculinities. Emphasizing the centrality of popular forms of cultural production to these processes and their potential critique, I examine practices of representation in transnational popular culture--particularly film, contemporary fiction, advertising, and television-to locate the making and unmaking of ideal transnational masculinities and their corresponding structures of social power. The individual chapters of the dissertation explore various forms of identitarian suicide within the triangulated transnational context I use to frame my analysis. My first chapter explores cultural responses to farmer suicide in India and its relationship to US-led practices of globalization. Next, moving away from the discussion of masculinities in ascent under structures of liberalization, I examine changing structures of white masculinity during the American recession, paying particular attention to representations of white male suicide in recent popular television. Building off of this, my third chapter examines Muslim American men in popular culture and the ways anxieties over suicide terrorism shape their representation. Building off my discussion on the American recession, this chapter further examines the changing states of proper/improper belonging in an American context increasingly defined by its global relationships. My final chapter investigates more figurative forms of suicide. This discussion explores the perceived willing devaluation of life among male migrant Indian laborers in the Persian Gulf States and how this experience of social death challenges established modes of belonging. Ultimately, this project challenges accepted notions of identity within the field of transnational American Studies by revealing the subversive formations found within unrecognized modes of self-representation.

Latent feature models for dyadic prediction

Menon, Aditya Krishna University of California, San Diego 2013 해외박사(DDOD)

소속기관이 구독 중이 아닌 경우 오후 4시부터 익일 오전 9시까지 원문보기가 가능합니다.

Following the Netflix prize, the collaborative filtering problem has gained significant attention within machine learning, spawning novel models and theoretical analyses. In parallel, the growth of social media has driven research in link prediction, with the aim of determining whether two individuals in a network are likely to know each other. Both problems involve the prediction of label (star ratings or friendship) between a pair of entities (user-movie or user-user). We call this general problem dyadic prediction. The problem arises in several other guises: predicting student responses to test questions, military disputes between nations, and clickthrough rates of webpages on ads, to name a few. In general, each such domain employs a markedly different approach, obscuring the underlying similarity of the problems being solved. This dissertation aims to explore the use of a single general method, based on latent feature modelling, for generic dyadic prediction problems. To this end, we make three contributions. First, we propose a generic <italic>framework</italic> with which to analyze dyadic prediction problems. This lets one reason about seemingly disparate problems in a unified manner. Second, we propose a <italic>model</italic> based on the log-linear framework, which is applicable to each of the aforementioned problems. The model learns <italic>latent features </italic> from dyadic data, and estimates a probability distribution over labels. Third, we systematically explore <italic>applications</italic> of our latent feature model to domains such as collaborative filtering, link prediction, and clickthrough rate prediction. In all cases, we show performance comparable or superior to existing state-of-the-art methods. For clickthrough rate prediction, ours represents the first application of latent feature modelling to the problem, demonstrating the value in a single framework with which to reason about these problems. We also show that latent feature modelling is scalable to datasets with hundreds of millions of observations on a single machine (the Netflix prize dataset), and hundreds of billions of observations on a small cluster (Yahoo! ad click data). We conclude with a discussion of future research directions, including transferring information from one network to another, and adapting to domains with extreme label sparsity.

The valuation of internal versus external knowledge: How competition within organizations focuses knowledge search beyond the organizational boundary

Menon, Tanya Stanford University 2000 해외박사(DDOD)

소속기관이 구독 중이 아닌 경우 오후 4시부터 익일 오전 9시까지 원문보기가 가능합니다.

This dissertation compares how managers value knowledge from sources within and outside the organizational boundary. Although much organizational and social psychological theory assumes that people favor internal knowledge, much empirical research has illustrated the prevalence of external knowledge transfer and the difficulties of internal knowledge transfer. An initial case study of a merger illustrates how one firm perceived the knowledge possessed by its acquisition more favorably when that firm was an external competitor as compared to within its boundary. Several surveys test the propositions derived from this case and indicate that some of the difficulties of valuing internal knowledge arise from its more negative social implications as compared to external knowledge. Managers expected more credit for using external knowledge as compared to internal knowledge and made more positive attributions about employees who used external knowledge; further, internally competitive organizational contexts promoted the valuation of external knowledge. A second case demonstrated how internal competition caused a firm to reject a technology when insiders possessed it yet embrace it once outsiders developed it. The evidence from these studies suggests that the social implications of knowledge use direct the patterns through which managers engage in the search for knowledge.

Geometric methods for the Maxwell-Bloch equations and the kinetics of martensitic phase transitions

Menon, Govind Brown University 2001 해외박사(DDOD)

소속기관이 구독 중이 아닌 경우 오후 4시부터 익일 오전 9시까지 원문보기가 가능합니다.

This dissertation has two parts. In both parts we use geometric methods to study dynamical systems with two scales. In the first part we study the Maxwell-Bloch equations for a two-level laser in a ring cavity. For Class A lasers, these equations have two distinct time scales, and form a singularly perturbed, semilinear hyperbolic system with two distinct characteristics. We extend Fenichel's geometric singular perturbation theory [40] to infinite dimensions by proving the persistence of a smooth, slow manifold under an unbounded perturbation. The proof relies on the energy preserving nature of the nonlinearity, and the existence of two characteristics. The slow manifold is a globally attracting, positively invariant manifold, with infinite dimension and codimension, that contains the attractor of the system. We rigorously decouple the slow and fast time scales) and obtain a reduced (but still infinite dimensional) dynamical system described by a functional differential equation. We also present the results of numerical computations. These demonstrate the applicability of our analysis and reveal a new type of spatiotemporal chaos in the Maxwell-Bloch equations. Independent of these scaling assumptions, we prove that the attractor of these equations has Gevrey regularity. Finally, we prove similar invariant manifold theorems for a class of infinite dimensional dynamical systems with relaxation. In the second part we study the weak limits of gradient dynamical systems with two spatial scales. These are higher dimensional generalisations of a model for the kinetics of martensitic phase transitions proposed by Abeyaratne, Chu and James [2]. We derive averaged equations in certain regions of phase space, but these equations typically do not have unique solutions. For the two dimensional problem we find that generically the phase space breaks into a countable number of domains, in the interior of which the homogenized dynamics are rectilinear. These domains have a Cantor set structure caused by the bifurcations of circle maps. Consequently, the homogenized equations vary on all scales, and we consider the implications for such models in materials science. V. P. Smyshlyaev has studied this problem independently, and some of our results are similar.