1,720,981 research outputs found
Learning analytics dashboards for adaptive support in face-to-face collaborative argumentation
No full textDespite the potential of learning analytics for personalized learning, it is seldom used to support collaborative learning particularly in face-to-face (F2F) learning contexts. This study uses learning analytics to develop a dashboard system that provides adaptive support for F2F collaborative argumentation (FCA). This study developed two dashboards for students and instructors, which enabled students to monitor their FCA process through adaptive feedback and helped the instructor provide adaptive support at the right time. The effectiveness of the dashboards was examined in a university class with 88 students (56 females, 32 males) for 4 weeks. The dashboards significantly improved the FCA process and outcomes, encouraging students to actively participate in FCA and create high-quality arguments. Students had a positive attitude toward the dashboard and perceived it as useful and easy to use. These findings indicate the usefulness of learning analytics dashboards in improving collaborative learning through adaptive feedback and support. Suggestions are provided on how to design dashboards for adaptive support in F2F learning contexts using learning analytics.Y
Analysis of Risk Factor Dependency Applying Data Mining Techniques to Traffic Accident Data
No full textChurn prediction of mobile and online casual games using play log data.
No full textInternet-connected devices, especially mobile devices such as smartphones, have become widely accessible in the past decade. Interaction with such devices has evolved into frequent and short-duration usage, and this phenomenon has resulted in a pervasive popularity of casual games in the game sector. On the other hand, development of casual games has become easier than ever as a result of the advancement of development tools. With the resulting fierce competition, now both acquisition and retention of users are the prime concerns in the field. In this study, we focus on churn prediction of mobile and online casual games. While churn prediction and analysis can provide important insights and action cues on retention, its application using play log data has been primitive or very limited in the casual game area. Most of the existing methods cannot be applied to casual games because casual game players tend to churn very quickly and they do not pay periodic subscription fees. Therefore, we focus on the new players and formally define churn using observation period (OP) and churn prediction period (CP). Using the definition, we develop a standard churn analysis process for casual games. We cover essential topics such as pre-processing of raw data, feature engineering including feature analysis, churn prediction modeling using traditional machine learning algorithms (logistic regression, gradient boosting, and random forests) and two deep learning algorithms (CNN and LSTM), and sensitivity analysis for OP and CP. Play log data of three different casual games are considered by analyzing a total of 193,443 unique player records and 10,874,958 play log records. While the analysis results provide useful insights, the overall results indicate that a small number of well-chosen features used as performance metrics might be sufficient for making important action decisions and that OP and CP should be properly chosen depending on the analysis goal
Data Requirements for Applying Machine Learning to Energy Disaggregation
No full textEnergy disaggregation, or nonintrusive load monitoring (NILM), is a technology for separating a household’s aggregate electricity consumption information. Although this technology was developed in 1992, its practical usage and mass deployment have been rather limited, possibly because the commonly used datasets are not adequate for NILM research. In this study, we report the findings from a newly collected dataset that contains 10 Hz sampling data for 58 houses. The dataset not only contains the aggregate measurements, but also individual appliance measurements for three types of appliances. By applying three classification algorithms (vanilla DNN (Deep Neural Network), ML (Machine Learning) with feature engineering, and CNN (Convolutional Neural Network) with hyper-parameter tuning) and a recent regression algorithm (Subtask Gated Network) to the new dataset, we show that NILM performance can be significantly limited when the data sampling rate is too low or when the number of distinct houses in the dataset is too small. The well-known NILM datasets that are popular in the research community do not meet these requirements. Our results indicate that higher quality datasets should be used to expedite the progress of NILM research
Short-Term Traffic Prediction With Deep Neural Networks: A Survey
No full textIn modern transportation systems, an enormous amount of traffic data is generated every day. This has led to rapid progress in short-term traffic prediction (STTP), in which deep learning methods have recently been applied. In traffic networks with complex spatiotemporal relationships, deep neural networks (DNNs) often perform well because they are capable of automatically extracting the most important features and patterns. In this study, we survey recent STTP studies applying deep networks from four perspectives. 1) We summarize input data representation methods according to the number and type of spatial and temporal dependencies involved. 2) We briefly explain a wide range of DNN techniques from the earliest networks, including Restricted Boltzmann Machines, to the most recent, including graph-based and meta-learning networks. 3) We summarize previous STTP studies in terms of the type of DNN techniques, application area, dataset and code availability, and the type of the represented spatiotemporal dependencies. 4) We compile public traffic datasets that are popular and can be used as the standard benchmarks. Finally, we suggest challenging issues and possible future research directions in STTP.
Who Should Participate in DR Program?Modeling with Machine Learning and Credit Scoring
No full textIn this study, we consider a residential DR program of incentive-based peak power reduction where invitation for participation can be sent selectively. The selective process can be crucial for improving efficiency of the program for two reasons. First, there are customers who do not change their behavior at all but take rewards due to the natural variations in their life patterns. Second, too many notifications can cause adversarial effects where participants turn off the notification channels or make complaint calls. For the selective process, obviously the process needs to be made as efficient as possible, but it is also essential to maximize the explainability of the selection process such that the operation of the program can be made smooth. To address this problem, we propose a customer participation behavior prediction model considering both accuracy and explainability, where the accuracy advantage of Machine Learning (ML) and the explainability advantage of Credit Scoring (CS) are combined. For the study, data was collected from 15,091 households in Korea for one year in 2016. ML algorithms, with up to 56 features, were studied and showed a fairly high prediction performance (AUROC 0.9576), but they were too complicated to satisfy explainability. A CS method of classing with a scorecard was adopted, where its explainability has been heavily tested and proven in the financial sector already. Direct adoption of general CS, however, does not guarantee an acceptable accuracy performance because energy data is quite different from financial data. To this end, we define a modified CS method using general CS as the base but with additional rules for high prediction performance. While this modified CS method maintains its explainability via a well-defined scorecard, it also shows comparable prediction performance as ML’s (AUROC 0.9509). The modified CS method is expected to affect residential DR in a positive way. Its high accuracy for predicting customer participation behavior means a large potential for improving efficiency. Its explainability means not only an easier interaction with customers but also less effort for educating call-center agents who need to deal with the customers
Utilizing Online Activity Data to Improve Face-to-Face Collaborative Learning in Technology-Enhanced Learning Environments
Full text link학위논문 (박사)-- 서울대학교 대학원 : 융합과학기술대학원 융합과학부(디지털정보융합전공), 2019. 2. Rhee, Wonjong .We live in a flood of information and face more and more complex problems that are difficult to be solved by a single individual. Collaboration with others is necessary to solve these problems. In educational practice, this leads to more attention on collaborative learning. Collaborative learning is a problem-solving process where students learn and work together with other peers to accomplish shared tasks. Through this group-based learning, students can develop collaborative problem-solving skills and improve the core competencies such as communication skills. However, there are many issues for collaborative learning to succeed, especially in a face-to-face learning environment. For example, group formation, the first step to design successful collaborative learning, requires a lot of time and effort. In addition, it is difficult for a small number of instructors to manage a large number of student groups when trying to monitor and support their learning process. These issues can amount hindrance to the effectiveness of face-to-face collaborative learning.
The purpose of this dissertation is to enhance the effectiveness of face-to-face collaborative learning with online activity data. First, online activity data is explored to find whether it can capture relevant student characteristics for group formation. If meaningful characteristics can be captured from the data, the entire group formation process can be performed more efficiently because the task can be automated. Second, learning analytics dashboards are implemented to provide adaptive support during a class. The dashboards system would monitor each group's collaboration status by utilizing online activity data that is collected during class in real-time, and provide adaptive feedback according to the status. Lastly, a predictive model is built to detect at-risk groups by utilizing the online activity data. The model is trained based on various features that represent important learning behaviors of a collaboration group.
The results reveal that online activity data can be utilized to address some of the issues we have in face-to-face collaborative learning. Student characteristics captured from the online activity data determined important group characteristics that significantly influenced group achievement. This indicates that student groups can be formed efficiently by utilizing the online activity data. In addition, the adaptive support provided by learning analytics dashboards significantly improved group process as well as achievement. Because the data allowed the dashboards system to monitor current learning status, appropriate feedback could be provided accordingly. This led to an improvement of both learning process and outcome. Finally, the predictive model could detect at-risk groups with high accuracy during the class. The random forest algorithm revealed important learning behaviors of a collaboration group that instructors should pay more attention to. The findings indicate that the online activity data can be utilized to address practical issues of face-to-face collaborative learning and to improve the group-based learning where the data is available.
Based on the investigation results, this dissertation makes contributions to learning analytics research and face-to-face collaborative learning in technology-enhanced learning environments. First, it can provide a concrete case study and a guide for future research that may take a learning analytics approach and utilize student activity data. Second, it adds a research endeavor to address challenges in face-to-face collaborative learning, which can lead to substantial enhancement of learning in educational practice. Third, it suggests interdisciplinary problem-solving approaches that can be applied to the real classroom context where online activity data is increasingly available with advanced technologies.Abstract i
Chapter 1. Introduction １
1.1. Motivation １
1.2. Research questions ４
1.3. Organization ６
Chapter 2. Background ８
2.1. Learning analytics ８
2.2. Collaborative learning ２２
2.3. Technology-enhanced learning environment ２７
Chapter 3. Heterogeneous group formation with online activity data ３５
3.1. Student characteristics for heterogeneous group formation ３６
3.2. Method ４１
3.3. Results ５１
3.4. Discussion ５９
3.5. Summary ６４
Chapter 4. Real-time dashboard for adaptive feedback in face-to-face CSCL ６７
4.1. Theoretical background ７０
4.2. Dashboard characteristics ８１
4.3. Evaluation of the dashboard ９４
4.4. Discussion １０７
4.5. Summary １１４
Chapter 5. Real-time detection of at-risk groups in face-to-face CSCL １１８
5.1. Important learning behaviors of group in collaborative argumentation １１８
5.2. Method １２０
5.3. Model performance and influential features １２５
5.4. Discussion １２９
5.5. Summary １３２
Chapter 6. Conclusion １３４
Bibliography １４０Docto
Enhancing Attribute-Factorized Representations in Variational Autoencoder by Regularizing Multiple Mutual Information Elements
Full text link학위논문(석사)--서울대학교 대학원 :융합과학기술대학원 융합과학부(디지털정보융합전공),2019. 8. Rhee, Wonjong.Recently, there have been many studies on deep generative models that can learn representations of data and generate new samples. We consider learning representations of target attributes and representations of the other attributes and how to disentangle them in deep generative models, and introduce a new Variational AutoEncoder (VAE) based generative model named as MMVAE (Multiple Mutual information elements VAE). The objective function of MMVAE can enhance attribute-factorized representations by regularizing multiple mutual information elements. Specifically, we construct a framework that explicitly regularizes mutual information of each pair among attributes, attribute representations, and the other representations by adopting Mutual Information Neural Estimation (MINE, Belghazi et al., 2018). In the model, the objective function consists of an evidence lower bound and three mutual information regularizers. The formulation corresponds to a minimax game, where a group of parameters in autoencoder is optimized to minimize the objective function while another group in mutual information regularizers is optimized to maximize the objective function. We demonstrate, through a series of experiments on CelebA datasets, that the model can learn the target attribute representations and the other representations in better factorized forms and that these factorized representations are useful for generating images with the target attributes.최근, 데이터의 표현 (representation)을 학습하고 새로운 샘플을 생성 할 수 있는 심층 생성 모델에 대한 연구가 활발하다. 우리는 특정한 속성 (attribute) 및 다른 속성과 관련된 표현들의 관계를 고려하여, 이들을 심층 생성 모델에서 어떻게 구분하여 처리할지에 대해 고찰하였다. 본 연구에서는 다수의 상호 정보량 (mutual information) 성분을 정규화하여 표현에서 속성의 요소분리를 강화시킬 수 있는, 변분법적 오토인코더 (Variational Autoencoder, VAE) 기반의 새로운 생성 모델 (MMVAE : Multiple mutual information VAE)과 목적 함수를 소개한다. 특히 Mutual Information Neural Estimation (MINE, Belghazi et al., 2018)을 채택하여, 속성의 레이블, 속성 표현 및 다른 표현 사이의 상호 정보량를 명시적으로 정규화하는 프레임 워크를 구성하였다. 이 모델에서 목적 함수는 증거 하한값 (evidence lower bound, ELBO)과 세 개의 상호 정보량으로 구성된다. 이는 미니맥스 게임 (mini-max game)에 해당하는데, 오토인코더의 매개 변수 그룹은 목적 함수를 최소화하도록 최적화되지만 상호 정보량의 매개 변수 그룹은 목적 함수를 최대화하도록 최적화 된다. 우리는 CelebA 데이터 세트에 대한 일련의 실험을 통해, MMVAE가 속성 표현과 다른 표현을 더 잘 구분하여 학습할 수 있고 이렇게 학습된 속성-요소분리된 표현 (attribute-factorized representation)은 주어진 속성을 포함하는 이미지를 생성하는 데 유용하다는 것을 입증하였다.I. Introduction 1
II. Related Works 8
2.1 VAE and CVAE 8
2.2 Recent Works for Attribute-Factorization 9
2.3 Mutual Information Neural Estimation 15
III. Research Questions and the Proposed Method 17
3.1 Research Questions 17
3.2 Model 18
3.2.1 ELBO 18
3.2.2 Mutual Information Regularization 21
3.2.3 Objective Function 23
3.3 Implementation 24
3.4 Evaluation methods 27
IV. Experimental Results 29
4.1 Experimental Setup 29
4.1.1 Dataset 29
4.1.2 Architecture of Neural Networks and Training 30
4.2 Experimental Results 33
4.2.1 Qualitative Results 33
4.2.2 Quantitative Results 36
V. Conclusion 42
5.1 Conclusions 42
5.2 Contributions 43
5.3 Limitations 43
References 46Maste
Deep Network Regularization with Representation Shaping
Full text link학위논문 (박사)-- 서울대학교 대학원 : 융합과학기술대학원 융합과학부(디지털정보융합전공), 2019. 2. Rhee, Wonjong.The statistical characteristics of learned representations such as correlation and representational sparsity are known to be relevant to the performance of deep learning methods. Also, learning meaningful and useful data representations by using regularization methods has been one of the central concerns in deep learning. In this dissertation, deep network regularization using representation shaping are studied. Roughly, the following questions are answered: what are the common statistical characteristics of representations that high-performing networks share? Do the characteristics have a causal relationship with performance? To answer the questions, five representation regularizers are proposed: class-wise Covariance Regularizer (cw-CR), Variance Regularizer (VR), class-wise Variance Regularizer (cw-VR), Rank Regularizer (RR), and class-wise Rank Regularizer (cw-RR). Significant performance improvements were found for a variety of tasks over popular benchmark datasets with the regularizers. The visualization of learned representations shows that the regularizers used in this work indeed perform distinct representation shaping. Then, with a variety of representation regularizers, a few statistical characteristics of learned representations including covariance, correlation, sparsity, dead unit, and rank are investigated. Our theoretical analysis and experimental results indicate that all the statistical characteristics considered in this work fail to show any general or causal pattern for improving performance. Mutual information I(zx) and I(zy) are examined as well, and it is shown that regularizers can affect I(zx) and thus indirectly influence the performance. Finally, two practical ways of using representation regularizers are presented to address the usefulness of representation regularizers: using a set of representation regularizers as a performance tuning tool and enhancing network compression with representation regularizers.Chapter 1. Introduction 1
1.1 Background and Motivation . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
Chapter 2. Generalization, Regularization, and Representation in Deep Learning 8
2.1 Deep Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2 Generalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2.1 Capacity, Overfitting, and Generalization . . . . . . . . . . . 11
2.2.2 Generalization in Deep Learning . . . . . . . . . . . . . . . . 12
2.3 Regularization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3.1 Capacity Control and Regularization . . . . . . . . . . . . . . 14
2.3.2 Regularization for Deep Learning . . . . . . . . . . . . . . . 16
2.4 Representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.4.1 Representation Learning . . . . . . . . . . . . . . . . . . . . 18
2.4.2 Representation Shaping . . . . . . . . . . . . . . . . . . . . 20
Chapter 3. Representation Regularizer Design with Class Information 26
3.1 Class-wise Representation Regularizers: cw-CR and cw-VR . . . . . 27
3.1.1 Basic Statistics of Representations . . . . . . . . . . . . . . . 27
3.1.2 cw-CR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.1.3 cw-VR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.1.4 Penalty Loss Functions and Gradients . . . . . . . . . . . . . 30
3.2 Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.2.1 Image Classification Task . . . . . . . . . . . . . . . . . . . 33
3.2.2 Image Reconstruction Task . . . . . . . . . . . . . . . . . . . 36
3.3 Analysis of Representation Characteristics . . . . . . . . . . . . . . . 36
3.3.1 Visualization . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.3.2 Quantitative Analysis . . . . . . . . . . . . . . . . . . . . . . 37
3.4 Layer Dependency . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
Chapter 4. Representation Characteristics and Their Relationship with Performance 42
4.1 Representation Characteristics . . . . . . . . . . . . . . . . . . . . . 43
4.2 Experimental Results of Representation Regularization . . . . . . . . 46
4.3 Scaling, Permutation, Covariance, and Correlation . . . . . . . . . . . 48
4.3.1 Identical Output Network (ION) . . . . . . . . . . . . . . . . 48
4.3.2 Possible Extensions for ION . . . . . . . . . . . . . . . . . . 51
4.4 Sparsity, Dead Unit, and Rank . . . . . . . . . . . . . . . . . . . . . 55
4.4.1 Analytical Relationship . . . . . . . . . . . . . . . . . . . . . 55
4.4.2 Rank Regularizer . . . . . . . . . . . . . . . . . . . . . . . . 56
4.4.3 A Controlled Experiment on Data Generation Process . . . . 58
4.5 Mutual Information . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
Chapter 5. Practical Ways of Using Representation Regularizers 65
5.1 Tuning Deep Network Performance Using Representation Regularizers 65
5.1.1 Experimental Settings and Conditions . . . . . . . . . . . . . 66
5.1.2 Consistently Well-performing Regularizer . . . . . . . . . . . 67
5.1.3 Performance Improvement Using Regularizers as a Set . . . . 68
5.2 Enhancing Network Compression Using Representation Regularizers 68
5.2.1 The Need for Network Compression . . . . . . . . . . . . . . 72
5.2.2 Three Typical Approaches for Network Compression . . . . . 73
5.2.3 Proposed Approaches and Experimental Results . . . . . . . 74
Chapter 6. Discussion 79
6.1 Implication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
6.1.1 Usefulness of Class Information . . . . . . . . . . . . . . . . 79
6.1.2 Comparison with Non-penalty Regularizers: Dropout and Batch Normalization . . . . . . . . . . . . . . . . . . . . . . . . . 81
6.1.3 Identical Output Network . . . . . . . . . . . . . . . . . . . 82
6.1.4 Using Representation Regularizers for Performance Tuning . 82
6.1.5 Benefits and Drawbacks of Different Statistical Characteristics of Representations . . . . . . . . . . . . . . . . . . . . . . . 83
6.2 Limitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
6.2.1 Understanding the Underlying Mechanism of Representation Regularization . . . . . . . . . . . . . . . . . . . . . . . . . 85
6.2.2 Manipulating Representation Characteristics other than Covariance and Variance for ReLU Networks . . . . . . . . . . . . 86
6.2.3 Investigating Representation Characteristics of Complicated Tasks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
6.3 Possible Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . 88
6.3.1 Interpreting Learned Representations via Visualization . . . . 88
6.3.2 Designing a Regularizer Utilizing Mutual Information . . . . 89
6.3.3 Applying Multiple Representation Regularizers to a Network . 90
6.3.4 Enhancing Deep Network Compression via Representation Manipulation . . . . . . . . . . . . . . . . . . . . . . . . . . 92
Chapter 7. Conclusion 93
Bibliography 94
Appendix 103
A Principal Component Analysis of Learned Representations . . . . . . 104
B Proofs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
Acknowlegement 113Docto
- …
