In this thesis, I present two instantiations of model parsimony for large scale object detection and discovery. For model inference, I present sparselet models which significantly reduce model inference complexity by utilizing a shared representation, reconstruction sparsity, and parallelism to enable real-time multiclass object detection with deformable part models at 5Hz with almost no decrease in task performance. For model learning, I present a framework for training object detectors using only one-bit image level annotations of object presence without any instance level annotations (i.e. bounding boxes).

关键词：对象检测；对象识别；机器学习；简约法

查看摘要 索取原文[62页]

Tracking travel patterns and modes is a goal that is useful on many levels, including calculating the transportation emissions of a population. Prior efforts to collect this information have been stymied by low accuracies or reliance on supplementary devices. In this paper, we describe a system that improves accuracy by using prompted recall on the smartphone, and aggregates the information to help detect large scale patterns. We also present the evaluation of a prototype implementation that was used to collect data from 44 users in the San Francisco Bay Area over 3 months.

关键词：交通排放；智能手机；自动化计算；量化模型

查看摘要 索取原文[15页]

This thesis introduces new techniques for specifying layout and generating efficient layout engines.First, we present a new programming methodology which addresses t he two central sources of bugs in layouts: ambiguities and conflicts. Then, we introduce a language of constraints in which we capture layout specifications formally. Finally, we show how to generate efficient layout engines automatically.

关键词：布局设计；布局引擎；可视化；编程

查看摘要 索取原文[98页]

We present a framework to design and verify the behavior of stochastic systems whose parameters are not known with certainty but are instead affected by modeling uncertainties, due for example to modeling errors, non-modeled dynamics or inaccuracies in the probability estimation. Our framework can be applied to the analysis of intrinsically randomized systems (e.g., random back off schemes in wireless protocols) and of abstractions of deterministic systems whose dynamics are interpreted stochastically to simplify their representation (e.g., the forecast of wind availability).

关键词：不确定性建模；概率系统；优化控制；形式技术

查看摘要 索取原文[210页]

We present a distributed-memory numerical library (Cyclops Tensor Framework (CTF)) that automatically manages tensor blocking and redistribution to perform any user-specified contractions. CTF serves as the distributed memory contraction engine in Aquarius, a new program designed for high-accuracy and massively-parallel quantum chemical computations. Aquarius implements a range of coupled-cluster and related methods such as CCSD and CCSDT by writing the equations on top of a C++ templated domain-specific language.

关键词：耦合簇运动方程；张量约缩；独眼巨人张量框架；CCSD

查看摘要 索取原文[28页]

This paper presents JITPROF, a profiling framework to dynamically identify code locations that prohibit profitable JIT optimizations. The basic idea is to associate execution counters with potentially JIT-unfriendly code locations and to use these counters to report code locations that match code patterns known to prohibit optimizations. We instantiate the idea for six JIT-unfriendly code patterns that cause performance problems in the Firefox and Chrome browsers, and we apply the approach to popular benchmark programs.

关键词：JavaScript；JIT；分析框架；定位

查看摘要 索取原文[15页]

This thesis presents two lines of work that fit into this theme. The first considers the approximation of the partition function, a pivotal quantity associated with spin systems. The second line of work is concerned with the complexity of exact computation of various natural mean observables of spin systems, e.g., the magnetization in the Ising model.

关键词：自旋系统；Ising模型；配分函数

查看摘要 索取原文[108页]

Breast cancer is a heterogeneous disease, consisting of at least five transcriptional subtypes described by distinct, but poorly understood, molecular profiles.

关键词：乳腺癌;异质性;细胞临床医学;疾病;分子;肿瘤;患者;型材;反应

查看摘要 索取原文[54页]

It is not clear why some prostate cancers are aggressive and progress to metastasis. Accumulating evidence suggests that the angiogenesis pathway may play a critical role. The significance of angiogenesis in prostate cancer is demonstrated by its correlation with Gleason score, clinical stage, progression, metastasis and survival. However, relatively few studies have assessed the role of genes involved in angiogenesis in recurrence of prostate cancer after radiotherapy. On the basis of strong biological rationale, we propose to comprehensively study this pathway in a well-characterized cohort of prostate cancer cases. Our hypothesis is that genetic and epigenetic individual variation in angiogenesis genes is associated with recurrence of prostate cancer after radiotherapy. We will test this hypothesis with a systematic evaluation of the 82 key genes in the angiogenesis pathway with recurrence of prostate cancer. The ultimate goal of this study is to identify biomarkers that can be used at the time of diagnosis to predict risk of recurrence and improve clinical treatment decision making.

关键词：遗传学;前列腺癌;放疗;血管生成;临床医学;基因;转移;前列腺;风险;测试和评估;示踪研究;复发

查看摘要 索取原文[7页]

Background. For diagnostic processes involving continual measurements from a single patient, conventional test characteristics, such as sensitivity and specificity, do not consider decision consistency, which might be a distinct, clinically relevant test characteristic.

关键词：诊断(医学);监测;患者;救护车;数据处理;决策;概率;比率;再版,测试和评估;创伤和损伤

查看摘要 索取原文[11页]