教学课程资源库 更多>>

关键词： Artificial intelligence   Bioinformatics   Computer science  

摘要： The Systems Biology community has invested a great deal of effort in modeling gene regulatory networks that should be able to (i) accurately predict future states and (ii) identify regulatory hubs that can be manipulated to achieve desired phenotypes. Most computational tools for the problem embody linear models (e.g. 5 * TF1 + 2 * TF2 - 0.4 * TF3....). However, it is well known that biological interactions are highly synergistic and non-linear. Further, those tools mostly try to directly predict networks even when the discovered edges (which usually come from some assay such as Chip-seq) may have little physiological significance (e.g., may not influence gene expression). This thesis considers an alternative approach to inferring gene causality. Specifically, we consider the problem of predicting the expression of genes at a future time point in a genomic time series. In this, we follow the philosophy that accurate prediction often corresponds to a good understanding of causality. The prediction may rest on several sources of data: the time point immediately preceding t, the entire target time series preceding t, and ancillary data. In biology, for example, the ancillary data may consist of a network based on binding data, data from different time series, steady state data, a community-blessed gold standard network, or some combination of those. We introduce OutPredict, which is a machine learning method for time series that incorporates ancillary steady state and network data to achieve a low error in gene expression prediction. We show that OutPredict outperforms several of the best state-of-the-art methods for prediction. The predictive models OutPredict in turn generate a causal network. Thus, this thesis presents an approach to the inference of causality based on predictions of out-of-sample time-points based on both steady state and time series data. Because the model for each gene identifies those transcription factors that have the most importance in predic

关键词： Computer science   Mathematics  

摘要： The purpose of this dissertation is to study two problems in combinatorial geometry in regard to obtaining better bounds on the number of geometric objects of interest: (i) monotone paths in geometric triangulations and (ii) pseudoline arrangements. (i) A directed path in a graph is monotone in direction of u if every edge in the path has a positive inner product with u. A path is monotone if it is monotone in some direction. Monotone paths are studied in optimization problems, specially in classical simplex algorithm in linear programming. We prove that the (maximum) number of monotone paths in a geometric triangulation of n points in the plane is O(1.7864n). This improves an earlier upper bound of O(1.8393n); the current best lower bound is Ω(1.7003n) (Dumitrescu et al., 2016). (ii) Arrangements of lines and pseudolines are fundamental objects in discrete and computational geometry. They also appear in other areas of computer science, for instance in the study of sorting networks. Let Bn be the number of nonisomorphic arrangements of n pseudolines and let bn = log2 Bn. The problem of estimating Bn was posed by Knuth in 1992. Knuth conjectured that bn ≤ (n/2) + o(n2) and also derived the first upper and lower bounds: bn ≤ 0.7924 (n2 + n) and bn ≥ n2/6 - O(n). The upper bound underwent several improvements, bn ≤ 0.6974 n2 (Felsner, 1997), and bn ≤ 0.6571 n2 (Felsner and Valtr, 2011), for large n. Here we show that bn ≥ cn2 - O(n log n) for some constant c > 0.2083. In particular, bn ≥ 0.2083 n2 for large n. This improves the previous best lower bound, bn ≥ 0.1887 n2, due to Felsner and Valtr (2011). Our arguments are elementary and geometric in nature. Further, our constructions are likely to spur new developments and improved lower bounds for related problems, such as in topological graph drawings. Developing efficient algorithms and computer search were key to verifying the validity of both results.

关键词： Ciphers   Security   Encryption   Maximum likelihood estimation   Light emitting diodes   Computer science   Wireless sensor networks   Ubiquitous sensor networks   lightweight ciphers   side-channel analysis   statistical fault analysis   Simeck  

摘要： With the development of wireless technology, the ubiquitous sensor networks have a profound effect on the way human interacts with computers, devices and environment. In order to reduce the potentially serious risks in the interaction, applying lightweight ciphers is effective to balance security, efficiency and convenience. Simeck is such a lightweight cipher that provides data confidentiality, authentication and integrity. It is significant to explore whether Simeck remains robust security. Up to now, the attacking assumptions of the previous security analysis of Simeck focus on the known-plaintext attack and the chosen-plaintext attack. There is no literature about Simeck against the ciphertext-only attack, which represents the weakest attacking capability of the attackers. On the assumption of the ciphertext-only attack, this paper proposes the security analysis of Simeck against the statistical fault analysis with a series of novel distinguishers of KDE, MME and MME-GF. The experimental results show that the proposed distinguishers can recover the secret key of Simeck in both decreasing faults and increasing reliability and accuracy. Thus, Simeck cannot resist against the statistical fault analysis with the proposed distinguishers. Furthermore, the good performance of these novel distinguishers can be applied on the PRESENT lightweight cipher. It offers the valuable reference for the design and analysis of the lightweight ciphers in the ubiquitous sensor networks.

关键词： Artificial intelligence   Design   Computer science  

摘要： All designs work within basic parameters and constraints to develop design solutions, however within urban housing developments current methodologies only consider a minimal required amount of inputs and data within a set amount of time. With no meaningful incentives to change development patterns affordable housing projects, along with their limited considerations, continue to create standardized, repetitive developments without any true responsiveness to the end user or surrounding context. We commonly see industries outside of architecture heavily influenced by data and technological advancements so much that it changes both companies internal operating structures and end-user products. This does not happen to nearly the same extent in the field of architecture. Digital technologies, such as generative design, allow architects to synthesize tremendous amounts of data and inputs within the same amount of time. Alternative project delivery methods, such as an integrated project delivery system, ensures that users values can be delivered within project constraints and promotes innovation throughout the entire process. By leaning in on computational capabilities and functioning under performance driven models, urban housing projects can begin to reflect aligned interests between users, builders, and architects. The possibility to consider more data in the form of both quantitative and qualitative values can allow designers to take responsibility and be more creative in developing urban housing solutions. Affordable housing demands continuously overburden new housing supply each year suggesting that the current system for developing affordable housing is more complex and restrictive than encouraging. Architects are limited to certain aspects where their ideas, advocacy’s, and decisions can influence change. Within the architectural aspects of means, methods, agreements, and products how can architects begin to influence change in the development of urban housing proje

关键词： Computer science   Artificial intelligence  

摘要： The abundance of personal health information available to healthcare professionals can be a facilitator to better care. However, it can also be a barrier, as the relevant information is often buried in the sheer amount of personal data, and healthcare professionals already lack time to take care of both patients and their data. This dissertation focuses on the role of natural language processing (NLP) in healthcare and how it can surface information relevant to healthcare professionals by modeling the extensive collections of documents that describe those whom they serve. In this dissertation, the extensive natural language data about a person is modeled as a set of documents, where the model inference is at the level of the individual, but evidence supporting that inference is found in a subset of their documents. The effectiveness of this modeling approach is demonstrated in the context of three healthcare applications. In the first application, clinical coding, document-level attention is used to model the hierarchy between a clinical encounter and its documents, jointly learning the encounter labels and the assignment of credits to specific documents. The second application, suicidality assessment using social media, further investigates how document-level attention can surface "high-signal" posts from the document set representing a potentially at-risk individual. Finally, the third application aims to help healthcare professionals write discharge summaries using an extract-then-abstract multidocument summarization pipeline to surface relevant information. As in many healthcare applications, these three applications seek to assist, not replace, clinicians. Evaluation and model design thus centers around healthcare professionals' needs. In clinical coding, document-level attention is shown to align well with professional clinical coders' expectations of evidence. In suicidality assessment, document-level attention leads to better and more time-efficient assessme

关键词： Aerospace engineering   Computer science   Engineering   Mathematics   Operations research   Remote sensing   Robotics   Transportation  

摘要： This thesis investigates the design of an autonomous navigation algorithm in GPS-denied environments. The performance of the current navigation solutions is dictated by the change in the operating environment. As a result, in the past two decades, various Simultaneous Localization and Mapping (SLAM) formulations have been proposed to adapt various operating conditions and improve robustness. Multi-sensor fusion has been one of the solutions. Every sensor has its own strengths and weaknesses to a given environment. By using multiple sensors, the weaknesses of one sensor can be counterbalanced by strengths of others. To this end, the proposed approach presents a novel tightly-coupled LIDAR-Inertial-Visual Odometry in a sliding window framework. Inertial measurement unit (IMU) provides high frequency inertial measurements, LIDAR provides an accurate map of surrounding, and visual odometry provides detailed information about the scene. Current state-of-the-art multi-sensor fusion works around coupling various odometry solutions in a pose-graph formulation. Such method is referred as loosely-coupled system where each sensor provides the robot pose estimate with uncertainty. However, a tightly-coupled system where all the sensor measurements are optimized in a factor-graph has been shown to achieve better accuracy and robustness compared to a loosely-coupled framework. In particular, the vision community has seen major improvements since adopting a tightly-coupled sensor fusion framework. This thesis investigates the performance of a tightly-coupled multi-sensor fusion in an optimization level. Further, inspired by the recently developed invariant extended Kalman filter (IEKF), the proposed odometry solution provides an accurate uncertainty estimate using matrix Lie group theory. The pipeline is validated, tested, and compared to the state-of-the-art SLAM algorithms.

关键词： Computer science   Statistics   Bayesian statistical decision theory   Spatial analysis (Statistics)  

摘要： This thesis discusses four novel statistical methods and approximate inference techniques for analyzing structured neural and molecular sequence data. The main contributions are new algorithms for approximate inference and learning in Bayesian latent variable models involving spatial statistics and nonlinear dynamics. First, we propose an amortized variational inference method to separate a set of overlapping signals into spatially localized source functions without knowledge of the original signals or the mixing process. In the second part of this dissertation, we discuss two approaches for uncovering nonlinear, smooth latent dynamics from sequential data. Both algorithms construct variational families on extensions of nonlinear state space models where the underlying systems are described by hidden stochastic differential equations. The first method proposes a structured approximate posterior describing spatially-dependent linear dynamics, as well as an algorithm that relies on the fixed-point iteration method to achieve convergence. The second method proposes a variational backward simulation technique from an unbiased estimate of the marginal likelihood defined through a subsampling process. In the final chapter, we develop connections between discrete and continuous variational sequential search for Bayesian phylogenetic inference. We propose a technique that uses sequential search to construct a variational objective defined on the composite space of non-clock phylogenetic trees. Each of these techniques are motivated by real problems within computational biology and applied to provide insights into the underlying structure of complex data.

关键词： Quantum physics   Condensed matter physics   Computer science  

摘要： Besides potentially delivering a huge leap in computational power, quantum computers also offer an essential platform for simulating properties of quantum systems. Consequently, various algorithms have been developed for approximating the dynamics of a target system on quantum computers. But generic quantum simulation algorithms—developed to simulate all Hamiltonians—are unlikely to result in optimal simulations of most physically relevant systems; optimal quantum algorithms need to exploit unique properties of target systems. The aim of this dissertation is to study two prominent properties of physical systems, namely locality and symmetry, and subsequently leverage these properties to design efficient quantum simulation algorithms. In the first part of the dissertation, we explore the locality of quantum systems and the fundamental limits on the propagation of information in power-law interacting systems. In particular, we prove upper limits on the speed at which information can propagate in power-law systems. We also demonstrate how such speed limits can be achieved by protocols for transferring quantum information and generating quantum entanglement. We then use these speed limits to constrain the propagation of error and improve the performance of digital quantum simulation. Additionally, we consider the implications of the speed limits on entanglement generation, the dynamics of correlation, the heating time, and the scrambling time in power-law interacting systems. In the second part of the dissertation, we propose a scheme to leverage the symmetry of target systems to suppress error in digital quantum simulation. We first study a phenomenon called destructive error interference, where the errors from different steps of a simulation cancel out each other. We then show that one can induce the destructive error interference by interweaving the simulation with unitary transformations generated by the symmetry of the target system, effectively providing a faster

摘要： Women are underrepresented in the STEM workforce areas of architecture, engineering, and computer science. The STEM workforce is a reflection of the STEM educational system. A solution for increasing the number of women in the STEM workforce is to increase the number of students in STEM education. There is research that explains why women do not pursue STEM in education and as a career but not why some do. The purpose of this qualitative study was to describe and explain the experiences and perceptions of young women who have graduated high school pertaining to their participation in and completion of an architecture, engineering, or computer science program. The research question used to guide this study was how do young women explain what influenced their entry into and completion of the Engineering, Architecture, or Computer Science pathway? Vroom’s expectancy theory was used as a theoretical lens to predict and classify the research findings. The participants described themselves as academically gifted and were influenced by their mother who worked in a STEM field. The young women who participated in the study entered the pathway without prior knowledge of what the pathway was about. They made the decision to stay and complete the pathway because they enjoyed the work and found their presence as one of the few young women in the class empowering.

关键词： Receivers & amplifiers   Software   Fourier transforms   Sensors   Signal processing   Antennas   Spacetime   Transmitters   False alarms   Algorithms   Visualization   Radar systems   Computer science   Electrical engineering   Engineering   Mathematics  

摘要： Target detection serve as one of the primary objectives in a radar system. From observations, contaminated by receiver thermal noise and interference, the processor needs to determine between target absence or target presence in the current measurements. To enable target detection, the observations are filtered by a series of signal processing algorithms. The algorithms aim to extract information used in subsequent calculations from the observations. In this thesis and the appended papers, we investigate two techniques used for radar signal processing; spectral estimation and space-time adaptive processing. In this thesis, spectral estimation is considered for signals that can be well represented by a parametric model. The considered problem aims to estimate frequency components and their corresponding amplitudes and damping factors from noisy measurements. In a radar system, the problem of gridless angle-Doppler-range estimation can be formulated in this way. The main contribution of our work includes an investigation of the connection between constraints on rank and matrix structure with the accuracy of the estimates. Space-time adaptive processing is a technique used to mitigate the influence of interference and receiver thermal noise in airborne radar systems. To obtain a proper mitigation, an accurate estimate of the space-time covariance matrix in the currently investigated cell under test is required. Such an estimate is based on secondary data from adjacent range bins to the cell under test. In this work, we consider airborne bistatic radar systems. Such systems obtains non-stationary secondary data due to geometry-induced range variations in the angle-Doppler domain. Thus, the secondary data will not follow the same distribution as the observed snapshot in the cell under test. In this work, we present a method which estimates the space-time covariance matrix based upon a parametric model of the current radar scenario. The parameters defining the scenario ar