教学课程资源库 更多>>

关键词： Distributed control systems   Physics   Deep learning   Credibility   Feedback   Software engineering   Sensors   Vehicles   Artificial intelligence   Computer science   Engineering   Transportation  

摘要： Context – The automotive industry is in the midst of a new reality where software is increasingly becoming the primary tool for delivering value to customers. While this has vastly improved their product offerings, vehicle manufacturers are increasingly facing the need to continuously develop, test, and deliver functionality, while maintaining high levels of quality. One important tool for achieving this is simulation-based testing where the external operating environment of a software system is simulated, enabling incremental development with rapid test feedback. However, the traditional practice of manually specifying simulation models for complex external environments involves immense engineering effort, while remaining vulnerable to inevitable assumptions and simplifications. Exploiting the increased availability of data that captures operational environments and scenarios from the field, this work takes a deep learning approach to train models that realistically simulate external environments, significantly increasing the credibility of simulation-driven software development. Contributions – First, focusing on simulating the input dependencies of automotive software functions, this work uses techniques of deep generative modeling to develop a framework for realistic test stimulus generation. Such models are trained self-supervised using recorded time-series field data and simulate the input environment much more credibly than manually specified models. With the credibility of stimulus generation being an important concern, an important concept of similarity as plausibility is introduced to evaluate the quality of generation during model training. Second, this work develops new techniques for sampling generative models that enable the controlled generation of test stimulus. Allowing testers to limit the range of scenarios considered for testing, the Metric-based Linear Interpolation (MLERP) sampling algorithm automatically chooses test stimuli that are verifiabl

关键词： Problem solving   Cognitive load   Software engineering   Computer science   Engineering   Mechanics   Physics  

摘要： A technical interview for software engineering roles consists of one or more stages within the interview life cycle. It begins with an initial screening of the candidate, usually conducted over the phone or through an online coding platform, such as CoderPad, Skype Interviews, and ***. Depending on their performance, the candidate may be invited for an on-site visit. This on-site visit consists of a series of one-on-one interviews (each 45 minutes to an hour, over a period of half a day to several days) with engineers, and sometimes managers. The interviews primarily focus on technical coding or algorithms, either on the whiteboard or using a simple text editor on a computer. During these sessions, candidates write code while thinking aloud as they work towards a solution, under the watchful eye of the interviewer(s). In other words, technical interviews are primarily a test of the candidates’ problem-solving or “analytical ability.” Despite the intended usage of these technical interviews, they can be notoriously unreliable method of assessing candidates’ qualifications. Identity-related aspects of a technical interview candidate (e.g., race, gender, disability, age) can affect interviewers’ behavior and the interview outcome. However, discrimination might not necessarily be the result of explicit bias on the interviewers’ part. An implicit bias may take place due to hidden factors in the existing technical interviews’ nature (e.g., breaking candidates’ thought process by asking them to think aloud or interrupting them.) These hidden factors may prevent certain candidates from performing properly despite their qualifications. The thesis of this dissertation is that the presence of an interviewer can disrupt a candidate’s problem-solving ability during a technical interview. Enhancing a candidate’s privacy can improve their problem-solving performance by reducing the stress arising from a social-evaluative threat. This dissertation intends to provide a t

关键词： Computer science   artificial intelligence   cybernetics  

ISBN: (数字)9788770226462 ISBN: (纸本)9788770226479

摘要： 5G, the emerging technology in mobile communication, is expected to deliver an important and decisive impact on several of the UN’s Sustainable Development Goals where universal accessibility to ICTs remains a serious concern. However, cyber security has emerged as a serious challenge, not least because of the increased accessibility and broader usage with associated vulnerability. Developing countries have additional challenges associated with both the expected faster build-up of accessibility and lack of qualified competencies within cyber security. Discussion of these challenges is the overall theme and motivation for this book. Technical topics discussed in the book include: • 5G in rural networks • Critical infrastructures • Open RAN • Protection of privacy • Cybersecurity and machine learning • Cybersecurity and disaster monitoring

关键词： 混合式教学   碎片化教学   计算机综合能力  

摘要： 基于案例分析，阐述利用“云班课”移动平台和线下课堂教学相结合的混合式教学形式，课程的碎片化教学，从而满足对计算机综合能力的应用型人才的需求。

关键词： Convex analysis   Algorithms   Computer science   Mathematics  

摘要： In this dissertation, we develop and analyze algorithms for robustness in three different machine learning settings. In the first part of the dissertation, we introduce the problem of hidden stratification — which is when a classification model substantially underperforms on certain unlabeled subclasses of the data — and propose a method to detect and mitigate this issue. Previous works studied how to address this in the setting where the subclass labels are known. Based on the empirical observation that unlabeled subclasses are often separable in the feature space of deep neural networks, we instead estimate subclass labels for the data using clustering techniques. We then use the estimated subclass labels as a form of noisy supervision in a distributionally robust optimization objective, in order to train a model that is more robust to inter-subclass variations. We demonstrate the effectiveness of our approach on several robust image classification benchmarks. We briefly discuss alternative methods for 1) utilizing a limited number of subclass labels to further improve performance, and 2) using contrastive learning to learn representations less susceptible to hidden stratification. In the second part of the dissertation, we study the problem of evaluating classification models under structured distribution shifts. Given a labeled sample from a "source" distribution and an unlabeled sample from the "target" distribution, importance weighting is the standard approach to perform such evaluations; however, importance weighting can struggle in high-dimensional settings, and fails when the support of the target distribution is not contained in that of the source. We show that one can sidestep these issues with some foreknowledge of the nature of the distribution shift; specifically, we present an algorithm that uses user-defined "slicing functions" — binary functions intended to capture possible axes of distribution shift — to estimate performance on the target distribu

关键词： Artificial intelligence   Computer science   Electrical engineering   Medical imaging  

关键词： Environment   Sustainability   Artificial intelligence   Computer science  

摘要： Artificial Intelligence (AI) has become an important area to tackle most environmental sustainability issues such as biodiversity, energy, transportation and water management. Biodiversity research has developed machine learning or natural language processing solutions to predict ecosystem services. Artificial intelligence applica-tions and machine learning models have been increasingly used for predicting and optimizing water resource conservation. Area neural network, expert systems, pattern recognition, and fuzzy logic models are the main focus areas in energy. Applications of computer vision and decision support were found in transportation. Timely monitoring of interventions is required to improve environmental sustainability.

关键词： Deep learning   Video recordings   Bandwidths   Image retrieval   Optimization   Energy   Algorithms   Queries   Performance evaluation   Servers   Computer science  

摘要： The widespread usage of mobile devices has led to the generation and storage of a large amount of videos and images. These images and videos are a rich source of information, and searching them for specific occurrences of actions and objects aids in various applications. Currently mobile devices have limited computational power to process some simple machine learning (ML) algorithms. Cellular and WiFi networking has enabled resource-constrained mobile devices to access powerful servers aided with graphical processing units (GPUs), and permitted offloading videos and images to be processed on the cloud. Various queries can be issued in mobile networks, and responding to such queries requires the consideration of the limited network bandwidth, and the computational and energy constraints on the devices. This thesis explores and investigates different techniques and mechanisms to address various queries in mobile networks. First, we address the problem of collecting images, stored on mobile devices, containing a queried object in a mobile network. The goal is to minimize the query response time given an energy constraint. In order to achieve this goal we divide the process into a pipeline with different stages. Each stage makes offloading decisions based on the network conditions, the cloud backlog and the hit-rate. A filtering stage is designed to run on the mobile device that uses a smaller less accurate but faster CNN to filter out images not containing the object of interest. Following that images that are positively classified in the filtering stage move to the selection stage where the more accurate CNN is used to classify the image and search for the queried object. At any stage, if the mobile device makes the decision to offload, the image is transmitted to the video cloud to be processed there. At the end, all the images that are positively classified should be uploaded to the server, where the requester can have access to them. Next, we investigate a differen

关键词： Earthquakes   Algorithms   Deformation   Volcanoes   Aftershocks   Geometry   Computer science   Geophysics   Mathematics   Mechanics   Physics  

摘要： In this thesis, we investigate the slow fault slip during continental seismic swarms and contribute to the understanding of slow fault slip by filling an identified observational gap for ruptures in the range between Mw 4 to 5.5 (Chapter 1). In the first core chapter (Chapter 2), we explore how to retrieve the fault geometry parameters from the interferometric phase observations during shallow seismic swarms. During shallow fault activity, a fault can rupture to the surface, which can cause phase discontinuities and affect the inversion. To overcome the phase unwrapping errors or surface faulting discontinuities, we propose to directly utilise the wrapped phase for geodetic inversion and develop new algorithms for data downsampling, weighting and misfit function construction. We find that this improved method has similar performance to retrieve models with respect to similar previous methods, but avoids the influence of phase unwrapping error. The chapter 3 presents a more realistic spatial distribution of fault slip based on the fault geometry retrieved with the method of Chapter 2. We utilise a new laboratoryderived crack model with finite shear stress at the crack tips and expand it to two dimensions. This new model uses few free parameters to describe the spatially variable fault slip distribution. This feature imposes an elliptical shape with asymmetric slip profile in our solutions, which acts as a different approach to regularise the geophysical inversion. The validity of this enhanced method is demonstrated by a compilation of published slip distribution models (Mw ≤ 7.5), and this indicates an elliptical crack can capture most of the fault slip distribution features, requiring lower degrees of freedom. The fourth chapter further explores the fault slip evolution in the time domain. We aim to retrieve the time-series of source parameters from non-simultaneous and temporally overlapped multi-sensor observations. To tackle numerical or numerical or noisy oscil

关键词： Task analysis   Semantics   Visualization   Monte Carlo methods   Supervised learning   Computer science   Annotations   Textbook question answering   multi-modality   diagram understanding  

摘要： Textbook Question Answering (TQA) is the task of answering diagram and non-diagram questions given large multi-modal contexts consisting of abundant text and diagrams. Deep text understandings and effective learning of diagram semantics are important for this task due to its specificity. In this paper, we propose a Weakly Supervised learning method for TQA (WSTQ), which regards the incompletely accurate results of essential intermediate procedures for this task as supervision to develop Text Matching (TM) and Relation Detection (RD) tasks and then employs the tasks to motivate itself to learn strong text comprehension and excellent diagram semantics respectively. Specifically, we apply the result of text retrieval to build positive as well as negative text pairs. In order to learn deep text understandings, we first pre-train the text understanding module of WSTQ on TM and then fine-tune it on TQA. We build positive as well as negative relation pairs by checking whether there is any overlap between the items/regions detected from diagrams using object detection. The RD task forces our method to learn the relationships between regions, which are crucial to express the diagram semantics. We train WSTQ on RD and TQA simultaneously, i.e., multitask learning, to obtain effective diagram semantics and then improve the TQA performance. Extensive experiments are carried out on CK12-QA and AI2D to verify the effectiveness of WSTQ. Experimental results show that our method achieves significant accuracy improvements of 5.02% and 4.12% on test splits of the above datasets respectively than the current state-of-the-art baseline. We have released our code on https://***/dr-majie/WSTQ.