教学课程资源库 更多>>

关键词： Civil engineering   Computer science   Climate Change   Artificial intelligence   Engineering   Fluid mechanics   Mechanics   Physics   Transportation  

摘要： Migration to urban areas is expected to approach 68% of the world population in 2050, according to UN estimates (Nations, 2018). Maintaining sustainable water distribution networks is imperative for transporting clean water to consumers, thereby ensuring public health. In addition, water distribution networks are essential infrastructures worldwide. Their structural safety is critical to ensure that treated water does not leak into the ground, wasting millions of tax dollars. Understanding the factors that affect the operational performance of a given water distribution system can help prioritize maintenance and predict the approximate service life of the pipelines such that replacement can be appropriately planned. Artificial intelligence (AI) for modeling and predicting the failure of water pipes has become advantageous in recent years. AI and machine learning approaches are fundamental, predictive models that help decision-makers develop strategies that mitigate the risk of failure by labeling pipes requiring immediate repair within a water distribution network. However, the failure process of water distribution pipelines remains ambiguous, and it may occur for ‘‘unknown reasons.’’ The intellectual contribution of this dissertation is to bridge the gap in the theoretical knowledge between critical factors and the deterioration of water distribution infrastructure. This dissertation also proposes new machine learning paradigms based on ensemble and deep learning to predict the failure of water distribution pipelines under various environmental and climatic conditions. To achieve the objectives of this dissertation, pipe failure data are collected from two municipalities in Canada, the City of London and Sainte-Foy in London and Quebec, respectively. In addition, climate data are amassed from the Environment and Climate Change Canada (ECCC) for the cities mentioned above. This dissertation research uncovers the effects of essential factors affecting the failure pre

关键词： Computer science   Artificial intelligence  

摘要： The advancement of information technology in coming years will bring significant changes to the way sensitive data is processed. But the volume of generated data is rapidly growing worldwide. Technologies such as cloud computing, fog computing, and the Internet of things (IoT) will offer business service providers and consumers opportunities to obtain effective and efficient services as well as enhance their experiences and services; increased availability and higher-quality services via real-time data processing augment the potential for technology to add value to everyday experiences. This improves human life quality and easiness. As promising as these technological innovations, they are prone to security issues such as data integrity and data consistency. However, as with any computer system, these services are not without risks. There is the possibility that systems might be infiltrated by malicious transactions and, as a result, data could be corrupted, which is a cause for concern. Once an attacker damages a set of data items, the damage can spread through the database. When valid transactions read corrupted data, they can update other data items based on the value read. Given the sensitive nature of important data and the critical need to provide real-time access for decision-making, it is vital that any damage done by a malicious transaction and spread by valid transactions must be corrected immediately and accurately. In this research, we develop three different novel models for employing fog computing technology in critical systems such as healthcare, intelligent government system and critical infrastructure systems. In the first model, we present two sub-models for using fog computing in healthcare: an architecture using fog modules with heterogeneous data, and another using fog modules with homogeneous data. We propose a unique approach for each module to assess the damage caused by malicious transactions, so that original data may be recovered and affec

关键词： Integer programming   Dynamic programming   Optimization   Decision making   Equilibrium   Game theory   Decomposition   Algorithms   Expected values   Computer science   Operations research  

摘要： We consider the hierarchical decision making (HDM) which is a game-theoretical decision process in which single or multiple decision-makers (DMs) solve sequential interdependent optimization problems over single or multiple planning periods. Depending on the relationship between decision-makers, the mathematical structure of the various decision problems, and the sequence in which decisions are made, HDM models can be classified as bilevel programs, stochastic programs, and robust optimization. This dissertation addresses a multistage stochastic programming models and two bilevel models. We start with stochastic programming where decision-makers cooperatively work toward a common objective, which can be viewed as a single decision maker making sequential decisions. We propose a reformulation for general multistage stochastic programming problems that are particularly amenable to parallel computing. The reformulation establishes bounds by decomposing the scenario tree into independent subtrees. Unlike traditional decomposition methods that solely focus on the scenarios, the presented approach allows a more general decomposition pattern that, as evident by the computational results, enables the method to deal with much larger instances. The last two chapters analyze bilevel mixed-integer models in the context of product transitions. Unlike stochastic programming, bilevel programming deals with problems where decision-makers are organized in a hierarchy with their own objectives. We model the hierarchical, decentralized nature of product transition by a multi-follower interdependent mixed-integer bilevel program. We propose an exact solution approach based on the cut-and-column generation algorithm. We then generalize this model by considering multiple product divisions that each operates in a specific market segment and require multiple resources to carry out their operations. Each product divisions is responsible for the development and production of new products. In

关键词： Computer science   Artificial intelligence  

摘要： There have been many advances in the artificial intelligence field due to the emergence of deep learning and big data. In almost all sub-fields, artificial neural networks have reached or exceeded human-level performance. However, most of the models are not interpretable and they perform like a black box. As a result, it is hard to trust their decisions, especially in life and death scenarios. In recent years, there has been a movement toward creating explainable artificial intelligence, but most work to date has concentrated on image processing models, as it is easier for humans to perceive visual patterns. There has been little work in other fields like natural language processing. By making our machine learning models more explainable and interpretable, we can learn about their logic, optimize them by removing bias, overcome their limitations, and make them resistant against adversarial attacks. This research dissertation is concentrated on making deep learning models that handle textual data, more understandable, and also use these insights in order to boost their performance by overcoming some of the common limitations. In addition to that, we use this knowledge to target words for designing efficient and effective textual adversarial attacks.

关键词： Computer science   Biology   Bioinformatics   Genetics  

摘要： Advances in genome sequencing have accelerated the growth of sequenced genomes but at a cost in the quality of genome annotation. At the same time, computational analysis is widely used for protein annotation, but a dearth of experimental verification has contributed to inaccurate annotation as well as to annotation error propagation. Thus, a tool to help life scientists with accurate protein annotation would be useful. In this work we describe a website we have developed, the Protein Annotation Surveillance Site (PASS), which provides such a tool. This website has three main components: a database of homologous clusters with more than nine million protein sequences deduced from the representative genomes of bacteria, archaea, eukarya, and viruses, together with sequence information; a machine-learning software tool which periodically queries the UniprotKB database to determine whether protein function has been experimentally verified; and a query-able webpage where the FASTA headers of sequences from the cluster best matching an input sequence are returned. The user can choose from these sequences to create a sequence similarity network to assist in annotation or else use their expert knowledge to choose an annotation from the cluster sequences. Illustrations demonstrating use of this website are given, and the Protein Annotation Surveillance Site (PASS) can be accessed at https://***/.

关键词： Computer science   Artificial intelligence  

摘要： The Internet has provided humanity with many great benefits, but it has also introduced new risks and dangers. E-commerce and other web portals have become large industries with big data. Criminals and other bad actors constantly seek to exploit these web properties through web attacks. Being able to properly detect these web attacks is a crucial component in the overall cybersecurity landscape. Machine learning is one tool that can assist in detecting web attacks. However, properly using machine learning to detect web attacks does not come without its challenges. Classification algorithms can have difficulty with severe levels of class imbalance. Class imbalance occurs when one class label disproportionately outnumbers another class label. For example, in cybersecurity, it is common for the negative (normal) label to severely outnumber the positive (attack) label. Another difficulty encountered in machine learning is models can be complex, thus making it difficult for even subject matter experts to truly understand a model’s detection process. Moreover, it is important for practitioners to determine which input features to include or exclude in their models for optimal detection performance. This dissertation studies machine learning algorithms in detecting web attacks with big data. Severe class imbalance is a common problem in cybersecurity, and mainstream machine learning research does not sufficiently consider this with web attacks. Our research first investigates the problems associated with severe class imbalance and rarity. Rarity is an extreme form of class imbalance where the positive class suffers extremely low positive class count, thus making it difficult for the classifiers to discriminate. In reducing imbalance, we demonstrate random undersampling can effectively mitigate the class imbalance and rarity problems associated with web attacks. Furthermore, our research introduces a novel feature popularity technique which produces easier to understand mod

关键词： Agriculture   Aerospace engineering   Organic chemistry   Artificial intelligence   Chemistry   Computer science   Engineering   Robotics   Transportation  

摘要： The fourth industrial revolution has brought about tremendous advancements in various sectors of the economy including the agricultural domain. Aimed at improving food production and alleviating poverty, these technological advancements through precision agriculture has ushered in optimized agricultural processes, real-time analysis and monitoring of agricultural data. The detrimental effects of applying agrochemicals in large or hard-to-reach farmlands and the need to treat a specific class of weed with a particular herbicide for effective weed elimination gave rise to the necessity of this research work. This research study involved the real-time detection of weeds and selective spray of herbicides on broadleaf and grass weed by applying deep learning algorithms on an embedded system. The deep learning algorithm deployed was a convolutional neural network model re-trained through transfer learning on the ResNet50 model using the soybean weed dataset and evaluated with the random forest classifier. The proposed smart selective herbicide sprayer was developed by assembling a quadcopter kit with the sprayer module and embedded system incorporated. The convolutional neural network model trained yielded training and validation accuracies of 99.98% and 98.4% respectively thus outperforming the random forest classifier. Also, training and validation losses of 0.0039 and 0.0323 respectively were obtained giving a clear indication that the model was appropriate. The test conducted on the smart herbicide sprayer, weighing about 3kg, indicates that the proposed models was able to detect broadleaf and grass weeds accurately and selectively spray herbicides on the weeds in less than a second. The results obtained from this work create the opportunity for more research to be carried out in this field of precision agriculture as regards weed and pest detection.

关键词： Computer engineering   Computer science   Artificial intelligence  

摘要： Neural networks span a wide range of applications of industrial and commercial significance. Binary neural networks (BNN) are particularly effective in trading accuracy for performance, energy efficiency or hardware/software complexity. In this thesis, I demonstrate a spintronic, re-configurable in-memory BNN accelerator, affectionately named PIMBALL: Processing In Memory BNN AcceL(L)erator, which allows for massively parallel and energy efficient computation. PIMBALL is capable of being used as a standard spintronic memory (STT-MRAM) array and a computational substrate simultaneously. For evaluation, I test PIMBALL using multiple image classifiers and a genomics kernel. The simulation results show that PIMBALL is more energy efficient than alternative CPU, GPU, and FPGA based implementations while delivering higher throughput.

关键词： Hydrologic sciences   Computer science   Hydraulic engineering   Sustainability   Water resources management   Limnology   Meteorology  

摘要： The Yahara River Chain of Lakes (RCL), located in central Dane County, WI recently have experienced flooding (water quantity) and eutrophication (water quality), resulting in negative outcomes. The objective of this dissertation is to address water quantity, such as appropriate water levels and flood risk, and water quality at beaches and streams in the Yahara RCL. The results of this dissertation are intended to increase our understanding of one of the world’s well-known lake systems - Yahara River Chain of Lakes. Furthermore, better understanding of how to manage and improve the Yahara RCL with broader impacts in application to other lake systems throughout the world is considered. In Chapter 2, the state of the art Integrated Nowcast and Forecast Operation System (INFOS) is developed to provide water information for the Yahara RCL. The system infrastructure consists of a web portal to retrieve and display observations that are used to drive models under a high- performance computing server. Water level and flow information are obtained from a suite of models that directly simulates the RCL system. INFOS reliably and effectively models real-time reverse flows due to sustained wind forcings or seiches, and flow choking due to channel constriction. Water level planning scenarios provide insight for lake management to reduce floods under extreme rainfall events. Overall, INFOS provides reliable and timely water information for the RCL for sharing to the public, managers, and researchers. In Chapter 3, sustainability assessment of flood risks from environmental, social, and economic perspectives in a RCL is performed. The flood assessment framework consists of five components including compiling data, modeling, mapping, estimating loss, and estimating risk. The modeled economic impact analysis reveals that the Yahara RCL is susceptible to large loss (in USD) at high return periods and long storm durations but a large risk at low return periods and long durations. The

关键词： Dentistry   Computer science   Biomedical engineering   Artificial intelligence   Software   Investigations   Web sites   Neural networks   Digitization  

摘要： Hypothesis and Objective: To compare the accuracy of cephalometric landmark identification between Artificial Intelligence (AI) Deep Learning Convolutional Neural Networks YOLOv3 algorithm and Manually Traced (MT) group. Methods: AAOF Legacy Denver collection was used to obtain 110 cephalometric images for this study. Lateral cephalograms were digitized by orthodontic resident in Dolphin Imaging after intra- and inter reliability check. The same images were uploaded to AI software Ceppro DDH Inc. Cartesian system of coordinates with Sela as 0:0 was used to extract x and y coordinate for 16 cephalometric points: Nasion, A point, B point, Menton, Gonion, Upper incisor tip, Lower incisor tip, Upper incisor apex, Lower incisor apex, Anterior Nasal Spine, Posterior Nasal Spine, Pogonion, Pterigomaxillary fissure point, Basion, Articulare and Orbitale. The mean distances were assessed relative to the reference value of 2 mms. Student paired t-tests at significance level of 5 % were used to compare the mean differences in each of the X- and Y-components. SPSS (IBM-vs. 27.0) software was use for the data analysis. Results: The results showed that there is no statistical difference for 12 out of 16 points when analyzing absolute difference between MT and AI group. Success detection rate for AI within 2 mm while comparing MT and AI group was 75 % and 93% within 4 mm. Conclusions: AI could be considered a promising tool to facilitate cephalometric tracing process in routine clinical practice and in research settings. Funding: no funding. IRB and/or ACC Protocol #: 2019-1180. Notice of Determination: activity Does Not Represent Human Subjects Research.