教学课程资源库 更多>>

关键词： Aerospace engineering   Computer science  

摘要： As the presence of Unmanned Aircraft Systems (UASs) become more prominent today and in the future. They are used in a variety of ways to solve solutions for a variety of tasks. UASs that are battery-powered typically have a flight time of no more than 30 minutes. Some tasks make take the drone beyond visual line of sight (BVLOS). The approach taken within this paper is allocating a secondary flight computer onboard the UAS to calculate paths while the primary computer controls the aircraft and follows the path being generated. With a proper map of the environment and use of a path planning algorithm the safety of the aircraft can be increased in missions that are BVLOS. This thesis will cover the concepts of path planning algorithms and the development of a modified version of a popular path planning algorithm. Show simulations of comparison with other variations of path planning algorithms and software in the loop (SITL) simulations on a fixed-wing aircraft. It will also show this algorithm's results when implemented in flight tests onboard a fixed-wing and multi-rotor UAS.

关键词： Computer science   Optics  

关键词： Intubation   Patients   Motivation   Age   Sample variance   Mutation   Cardiac catheterization   Medical research   Clinical trials   Design   Biomarkers   Algorithms   Performance evaluation   Precision medicine   Computer science   Medicine   Pharmacology  

摘要： A key objective in an interventional study, such as a randomised clinical trial, is the evaluation of heterogeneity of treatment effect in the population. This allows us to identify the most promising intervention for a given observation. In this thesis we approach this by targeting two tightly coupled sub-problems. The first concerns the identification of covariates and the second the identification of subgroups associated with treatment effect heterogeneity. Regarding the first problem we study an information theoretic approach. This can be motivated by phrasing the predictive covariate selection problem in log-likelihood terms. We study the properties of this approach in the case of randomised studies and evaluate low-dimensional approximations that are better suited for small-sample and/or high-dimensional studies. We identify some limitations and propose extensions based on propensity score weighting and stratification that extend this criterion in scenarios when the treatment assignment depends on the covariates. Regarding the second problem, we discuss recursive partitioning approaches coupled with weighting methods for treatment effect estimation. The purpose of these methods is to tackle the problem of subgroup identification in the presence of confounders in the data. Finally, studying the literature of subgroup identification we identify a significant number of approaches. Given such a large number of methods to choose from, an important question is how to select the best for a given task. We introduce a framework that uses the subgroup stability as a measure to capture the variations in the identified subgroups due to small changes in the data.

关键词： Itemsets   Data mining   Production facilities   Databases   Memory management   Computer science   Production planning   Data mining   erasable itemset mining   bitmap representation   product database  

摘要： Erasable-itemset mining is a valuable method of pattern extraction for helping the manager of a factory analyze production planning. The erasable itemsets derived can be considered important production information regarding how to plan the production of a factory during an economic depression or financial shortage for the manager. After the erasable-itemset mining was proposed in 2009, several efficient mining approaches for finding erasable itemsets have been developed. However, these methods require a considerable amount of execution time when the amount of product data is large. Especially, manufacturing small amounts of versatile products has been a trend today, and it will generate a large product database. This paper adopts a bitmap representation for itemsets in an erasable-itemset mining algorithm to speed up the execution. Unlike the traditional bitmap meaning for frequent itemsets, a bitmap for erasable-itemset mining here denotes the relationship that a product includes at least one material (item) in a specified itemset. Using the bitmap representation can easily find the desired products to check, thus decreasing the scans of a database. Experimental evaluation on synthesized and real datasets was used to compare the proposed approach with the other two under different parameter values. The experimental results show that the proposed approach can make a good trade-off between execution time and memory usage.

关键词： Sparsity   Sentiment analysis   Optimization techniques   Knowledge   Pandemics   Support vector machines   Prayer   Earthquakes   Linear programming   Algorithms   Emotions   Coronaviruses   COVID-19   Computer science   Geophysics  

摘要： One aspect of crisis management involves the ability to understand the emotional reactions of people to adjust the response strategies. Uncovering the topics and emotions related to an event over time is critical for timely intervention. For fast-evolving events, microblogs tend to be the preferred means of communication for people to share information and feelings. In this thesis, we develop an event analysis framework to identify the topics and emotional reactions of people in microblogs. The framework has three components: (1) a trend interval detection algorithm to determine the granularity for discovering hot topics while minimizing potential information loss, (2) a weakly supervised learning approach to learn coherent topics and their corresponding emotional reactions, (3) an event summary generator that gives representative microblogs and emotion distributions associated with the topics over time. Extensive experiments on multiple real-world event datasets demonstrate the effectiveness of the proposed approach over existing solutions.

关键词： Computer science  

摘要： Natural Language Processing (NLP) tasks have witnessed a significant improvement in performance by utilizing the power of end-to-end neural network models. An NLP system built for one job can contribute to other closely related tasks. Coreference Resolution (CR) systems work on resolving references and are at the core of many NLP tasks. The coreference resolution refers to the linking of repeated object references in a text. CR systems can boost the performance of downstream NLP tasks, such as Text Summarization, Question Answering, Machine Translation, etc. We provide a detailed comparative error analysis of two state-of-the-art coreference resolution systems to understand error distribution in the predicted output. The understanding of error distribution is helpful to interpret the system behavior. Eventually, this will contribute to the selection of an optimal CR system for a specific target task.

关键词： Data models   Filtering   Computer science   Classification algorithms   Licenses   Interpolation   Distributed databases   Imbalanced learning   oversampling   SMOTE   radius distance   initial selection  

摘要： Imbalanced learning problems are a challenge faced by classifiers when data samples have an unbalanced distribution in each class. Furthermore, the synthetic oversampling method (SMOTE) is a preprocessing technique widely used to synthesize new data and balance the different numbers of samples in each class. One of the SMOTE method's expansions is based on the initial selection approach, which determines the best candidates to be oversampled in the data before the process of synthetic example generation starts. However, SMOTE and most of the existing oversampling methods based on initial selection still found overlapping data on the final result. This issue makes it difficult for any classifiers to determine the decision boundary of each class. Therefore, this research proposes a new oversampling technique called Radius-SMOTE, which emphasizes the initial selection approach by creating synthetic data based on a safe radius distance. Furthermore, new synthetic data are prevented from overlapping in the opposite class with the safe radius distance. The Radius-SMOTE was evaluated extensively with thirteen artificial imbalanced datasets from the KEEL repository. The experimental results show that the proposed method is able to achieve the best results on 5 datasets, namely yeast-1-4-5-8_vs_7, ecoli-0-1-3-7_vs_2-6, Umbilical cord, Pima, and Haberman dataset in term of various assessment metrics. Besides that, the computational cost for our proposed method is also relatively low, with an average time of 0.5 to 1 second on the 13 tested datasets.

关键词： Ankle   Gait   Kinematics   Running   Software   Motion capture   Range of motion   Muscle contraction   Kinetics   Force   Posture   Biomechanics   Computer science   Mechanics   Morphology   Physics  

摘要： INTRODUCTION: Coupling in the distal foot may be due, at least in part, to the foot's windlass mechanism. This mechanism has been demonstrated passively, but its role in dynamic movement is still unclear. A systematic manipulation of metatarsophalangeal (MTP) mechanics may help determine to what extent distal foot coupling during dynamic and active movement is due to the windlass mechanism versus active muscle contractions or springlike ligaments. Furthermore, exploring the windlass mechanism in feet with varying foot structure may aid our understanding of the relationship between foot structure and foot function. PURPOSE: The overall purpose of this study is to investigate the kinematic and kinetic coupling between the MTP and midtarsal joints through a systematic manipulation of the windlass mechanism (achieved through methodical changes to MTP motion). Additionally, we aimed to explore the relationship between foot structure and the efficacy of the windlass mechanism during passive, active, and dynamic movement. METHODS: First, arch height and flexibility were measured using the Arch Height Index Measurement System. Next, participants performed four order-randomized conditions where MTP extension was isolated: 1) Seated Passive MTP Extension, 2) Seated Active MTP Extension, 3) Standing Passive MTP Extension, and 4) Standing Active MTP Extension. Lastly, participants performed three heel raise conditions that manipulated the starting position of the MTP joint: 1) Neutral: normal heel raise, 2) ToeExt: heel raise with the toes placed on an inclined surface of 30 degrees to put the MTP joint into extension, and 3) ToeFlex: heel raise with the toes placed on a declined surface of 30 degrees to put the MTP joint into flexion. All conditions were performed to a metronome of 40 beats per minute to control angular velocity. A kinetic multisegment foot model was created in Visual 3D software and used to calculate ankle, midtarsal, and MTP joint angles, moments, powers, an

关键词： Computer science   Artificial intelligence   Computer engineering  

摘要： The growth of cyber attacks in both numbers and varieties in recent years demands building a more sophisticated network intrusion detection system (NIDS). Signature-based NIDS, which were traditionally being used to detect malicious traffic, can't cope with today's variety of network attacks, as they are incapable of detecting attacks that have not been seen before and, therefore, do not have a known signature to detect. As a result, Machine Learning-based NIDS built on neural networks are getting more attention due to their ability to detect such attacks. While moving towards ML-based NIDS holds promise, such systems have their own drawbacks. The first problem is the vulnerability of neural networks to adversarial examples. It has been shown that an adversary can fool deep learning models used in tasks such as image classification by applying small and imperceptible modifications to the input images to force the models to predict them as their desired class. But the existing ML-based NIDS were not evaluated in such an adversarial setting, even though their potential of being vulnerable to the adversarial examples can cause the network systems to face significant threats. Secondly, training a model in a supervised manner on network traffic has its own challenges. One major issue is that labeling a huge dataset consisting of billions of packets is expensive and time-consuming. Even if we can label such a dataset and train a model on it, the model can hardly detect new attacks not included in the train set. For this reason, existing work mostly trained their models in an unsupervised fashion, but in order to detect different network attacks, these models generate so many false alerts, and any attempt to reduce their false positive rate significantly deteriorates their detection rate. In this dissertation, we first identify the challenges to evaluate ML-based NIDS trained on network traffic in an adversarial setting. We demonstrate how an adversary can craft adversaria

关键词： Biology   Computer science   Conservation biology   Ecology   Mathematics  

摘要： Dispersal and interaction networks underlie many ecological systems, often dictating spatio-temporal dynamics as well as emergent patterns. Dispersal has implications for establishment for distinct subpopulations of individual species as well as for the emergence of universal spatial patterns and varied community level structures. Such implications and the search for mechanisms that drive them are the central themes of this thesis. The accompanying manuscripts discuss two very different approaches of analysing ecological systems. The less complex of the two systems consists of sites along a coastal landscape that harbours the seagrass Halodule uninervis. The task is to partition the entire population into subpopulations that have very little dispersal across them. We use an algorithm to aggregates sites that are more strongly connected in the same subpopulation for a range of dispersal scenarios. These scenarios are useful since the dispersal processes and life-spans of the seagrass are not well-understood. In subsequent work, the idea of complexity is more precise. We analyse ecosystems whose complexity is captured by the strength of interactions between species and the fraction of interaction links realized out of all possible ones. On one hand we study how ecological communities on islands are assembled from a large pool of species that can also immigrate from the mainland source, and if it can provide mechanisms to explain the scaling of species richness with island areas – namely the species-area relationship. A drastically different setting is large contiguous landscapes. We posit that species interactions vary across space, which helps characterize habitat patches that would be connected by high dispersal of species. This simplified picture provides a range of situations to understand how habitat heterogeneity affects species richness. An overarching theme in investigating these complex ecological systems is to first analyze the fully random case without assu