教学课程资源库 更多>>

关键词： Education   Computer science  

摘要： The inherently complex nature of evaluating teamwork calls for methods to measure and predict students’ performance and provide timely feedback. Analyzing students’ individual performance, particularly in low-stake teams, is a challenge since the main goal in such team settings is knowledge acquisition and social skill development rather than final artifact production. Students’ attitude plays an important role in their cognition process and is commonly measured by self-report tools or asynchronous communications in textual discussion platforms. These methods have certain drawbacks such as distracting the learning process, demanding time and commitment from students or lack of emotional awareness which reduces the reliability of such tools. Research suggests speech is the best way to measure attitude accurately since it captures behavior directly rather than self-report. This study focuses on operationalizing attitude constructs of affect, self-efficacy, and personality and analyzing their correlation with students’ performance in order to identify which attitude constructs serve as performance predictor metrics. Personality and self-efficacy are measured using standard self-report tools, and affective states are captured from students’ conversations as they work in low-stake teams in a CS1 active learning class and discuss course content with peers. The novelty of this research is its focus on students' speech in class and how to operationalize affective states from verbal conversations as a metric that is related to individual performance. We record students’ conversations during teamwork in multiple sessions throughout the semester. By applying Natural Language Processing (NLP) algorithms we conduct sentiment analysis to detect valence, polarity and multiple classes of affective states from the conversations. The result of data analysis shows that students with higher levels of positive sentiment during the semester had higher performance scores. The result of pe

关键词： Computer science   Diseases   Correlation   Genetic algorithms   Feature extraction   Robots   Object recognition  

摘要： The affliation of Kashif Javed was incorrectly listed in [1] ; it should be “Department of Robotics, SMME, NUST, Islamabad, Pakistan.”

关键词： Botany   Artificial intelligence   Computer science  

摘要： Automatic identification of plants from natural images is a challenging problem that is relevant to both the disciplines of Botany and Computer Science. The classification of plant images at the species level is a computer vision task called fine-grained categorization. This categorization problem is particularly complicated due to a large number of plant species, the inter-species similarity, the large-scale variation in appearance, and the lack of annotated data. Despite the availability of dozens of plant identification mobile applications, categorizing plant species from natural images remains an unsolved problem—e.g., most of the existing applications do not address the multi-scale nature of this type of image. Furthermore, an automated system capable of addressing the complexity of this computer vision problem has important implications for society at large, not only in preserving ecosystem biodiversity and public education but also in numerous agricultural activities such as detecting abnormalities in plants and analyzing food crops. In this dissertation, I present a new approach to the problem of automatically categorizing plant species using photos taken in nature. Essentially, this approach assembles a collection of Convolutional Neural Networks (CNN-based) to create a plant categorization system that I named WTPlant (What's That Plant?). One of the novelties of this system is a preprocessing method that extracts multi-scale samples from natural images, making the classification models more robust to variations in the scale of the plant. A comprehensive experimental evaluation of this new preprocessing method compares its performance with frequently used data augmentation techniques over different classification models of the system. WTPlant also enables the categorization of multiple plant components simultaneously by employing distinct classification pipelines for plants (leaves, branches, bushes, and trees) and flowers. The combination of these multi-or

关键词： Computer science  

摘要： Our society has become increasingly reliant on software systems; however, these same software systems are notoriously buggy. Furthermore, manually locating and debugging these faults is both time and resource intensive. Automated program repair techniques were created to mitigate this burden. Despite the relative popularity of these techniques, all suffer from scalability and accuracy limitations. Most notably, these techniques generate large search spaces of potential patches, bloating the runtime. The key novelty of our work is the creation of an automatic defect classification technique. Our vision is for this defect classification to become a new input for any automated program repair technique, which can then be leveraged to guide their patch generation subroutine, thus reducing the search space and improving scalability. We present an automated defect classification technique for programs written in C which utilizes machine learning. We address the following: (1) how to best classify defects in software (which/what metrics are best), (2) the effectiveness of different machine learning classification algorithms on the defect classes, (3) the run time and metrics compared with established software program, (4) how the classification method can be implemented in automated program repair, and (5) how valuable defect classification can be to secure software fixes and program analysis. The results of our classification technique presented challenges in accuracy ratings. The testing phase for the machine learning technique resulted in an average of 20% accuracy for all algorithms with higher results in the decision tree method of machine learning classification. The results motivate exploration of key future improvements including: manipulation of attribute classes, boosting methods, other machine learning techniques, and in-depth fixtures based on classes. While we focus on defect classification for C programs, our technique highlights the efficacy of developing suc

关键词： Robotics   Computer science  

摘要： When we are babies, we learn how to see by watching how the world changes and by interacting with it. Can we use these same signals to train vision models? In this thesis, we outline several works which use these paradigms as a basis for learning algorithms. First, we explore learning by watching in which video data is directly used to learn about the visual world. Second, we tackle multiple challenging tasks in embodied environments in which agents learn by interacting with their surroundings.

关键词： Computer engineering   Agriculture engineering   Artificial intelligence   Computer science  

摘要： Chlorophyll fluorescence (ChlF), being an integral part of the photosynthetic process, has long been known to be a useful tool to detect plant stress. Early and accurate plant stress detection is imperative in enabling timely and appropriate intervention. One of the major limitations of prior work for ChIF-based plant classification is that, in general, only a few key inflection points of a localized section of a chlorophyll fluorescence signal are used to calculate single index values. These values yield very limited insight into stress level or stress type. In this thesis, we propose two new methods for clustering or classifying plant stress based upon the use of Hidden Markov Models (HMMs) using global (versus local) ChlF time-varying signal data acquired via video imaging. The first is an unsupervised method for plant stress clustering. The second is a supervised method for plant stress classification. To the best of our knowledge this is the first use of HMMs in the field of plant stress detection. We first introduce and present the proposed unsupervised learning method. We cluster the time-varying-intensity-signal data using unsupervised learning via HMMs. We show how in some scenarios plant stress classification is improved by the selection of a low-pass filtered plant’s entire chlorophyll fluorescence signal profile, (as a global feature selection), the rate-of-change-in-time of the plant ChlF intensity time-varying profile, (as another global feature selection) and data quantization. We show that HMMs allow more variability in classifying data than other raw data distance-based metrics. We explore the ergodic and Bakis models for HMM state transition matrix initialization. In addition we propose a new method for initialization of the HMM state transition matrix, denoted as “the state information based initial probability assignment” (SIPA) method and compare it to the heuristic initial probability assignment (HIPA) method. We show how using the Bayesian inf

摘要： In this article the author shares his experience of reading Cheryl Thompson's book, 'Beauty in a Box: Detangling the Roots of Canada's Black Beauty Culture', which maps out Canada's Black beauty industry from past to present, infusing critical race theory and Black feminism into a historical investigation on Black hair; and mentions that availability of ethno-specific products and larger structures of racism are intrinsically connected.

关键词： Aerospace engineering   Robotics   Computer science  

摘要： We present a Kalman filter-based method for event-triggered decentralized data fusion in cooperative localization problems that can be scaled to large numbers of networked autonomous robots operating in a swarm. This scaling is accomplished without significantly increasing the number of states each robot must estimate locally or the amount of data transferred through the network, thanks to implicit measurement fusion updates afforded by event-triggered communication. To keep state estimates suitably synchronized in this event-triggered approach, we exploit a novel method for performing reduced order partial covariance intersection between local estimates, such that only the overlapping set of swarm states between any two communicating agents’ estimates need to be fused. This reduces the number of swarm states each agent must track to only the subset of swarm states that have a significant effect on that agent’s ‘ownship’ state estimates. This leads to significantly simpler onboard computation for each agent and reduces the amount of data that must be transferred through the network when synchronization and measurement messages are considered together. Cooperative localization simulations for a 30 robot swarm show that the method is able to achieve good localization performance, even for greatly reduced measurement sharing and greatly reduced overall data transfer compared to full state decentralized fusion. We then present an architecture for integrating our event-triggered decentralized data fusion method with navigation filtering to act as a proxy for GPS aiding measurements, in order to extend its application to robotic systems with complex dynamics such as those typically present in physical systems. Simulations in a 6 vehicle underwater scenario show that the method works to provide GPS-like aiding to the navigation filters, even when vehicles do not have access to GPS, all the while sharing less data than standard decentralized data fusion. Next, we use a quan

关键词： Artificial intelligence   Computer science  

摘要： Deep learning techniques have revolutionized many areas of science due to their unique properties like adaptability, scalability, and the potential to rapidly adjust to new and unknown challenges. This thesis aims to revolutionize the methodology behind the millimeter-wave (mmWave) beamforming in the same way. The main objective of this thesis is to reduce the time complexity for finding the best beam pair in an mmWave link by transitioning from a conventional brute force approach of testing all possible beam pairs, with polynomial time complexity of O(????2), to a deep learning solution, with time complexity of constant time of O(1). The collected SNR data of all beam pairs was interpreted to a multivariate parallel time series for the lateral motion with a fixed step size. However, the main challenge was to choose how to fit this data to an LSTM, a short form of Long Short Term Memory, considering the technical limitations to build numerous different multivariate time series models for each beam pair. Even if we somehow manage to build an independent LSTM model for each beam pair, then there would be very high chances of missing the inherent relationship among the beam pairs. Hence, we came up with an approach that combines both of the shortcomings - a multivariate parallel time series approach. To ensure that the data we used in our evaluation is reliable and an accurate representation of a real-world scenario, we collected the data using a real hardware – the X60 testbed. X60 is the first SDR-based testbed for 60 GHz WLANs, featuring fully programmable MAC/PHY/Network layers, multi-Gbps rates, and a user-configurable 12-element phased antenna array. Our evaluation shows that deep learning can considerably improve the prediction accuracy, and the model can achieve high throughput with little performance loss and with almost zero overhead.

关键词： Computer science  

摘要： Individuals with voice disorders experience challenges communicating daily. These challenges lead to a significant decrease in the quality of life for individuals with dysphonia. While voice amplification systems are often employed as a voice-assistive technology, individuals with voice disorders generally still experience difficulties being understood while using voice amplification systems. With the goal of developing systems that help improve the quality of life of individuals with dysphonia, this work outlines the landscape of voice-assistive technology, the inaccessibility of state-of-the-art voice-based technology and the need for the development of intelligibility improving voice-assistive technologies designed both with and for individuals with voice disorders. With the rise of voice-based technologies in society, in order for everyone to participate in the use of voice-based technologies individuals with voice disorders must be included in both the data that is used to train these systems and the design process. An important and necessary step towards the development of better voice assistive technology as well as more inclusive voice-based systems is the creation of a large, publicly available dataset of dysphonic speech. To this end, a web-based platform to crowdsource voice disorder speech was developed to create such a dataset. This dataset will be released so that it is freely and publicly available to stimulate research in the field of voice-assistive technologies. Future work includes building a robust intelligibility estimation model, as well as employing that model to measure, and therefore enhance, the intelligibility of a given utterance. The hope is that this model will lead to the development of voice-assistive technology using state-of-the-art machine learning models to help individuals with voice disorders be better understood.