Open Access. Powered by Scholars. Published by Universities.®
Physical Sciences and Mathematics Commons™
Open Access. Powered by Scholars. Published by Universities.®
- Institution
-
- Singapore Management University (145)
- Walden University (53)
- Selected Works (13)
- San Jose State University (11)
- California Polytechnic State University, San Luis Obispo (6)
-
- Old Dominion University (6)
- The University of Akron (6)
- Florida International University (5)
- Kennesaw State University (5)
- University of Massachusetts Amherst (5)
- University of Nebraska at Omaha (5)
- Wright State University (5)
- Dakota State University (4)
- The University of Maine (4)
- University for Business and Technology in Kosovo (4)
- University of South Florida (4)
- University of Wisconsin Milwaukee (4)
- Air Force Institute of Technology (3)
- City University of New York (CUNY) (3)
- Embry-Riddle Aeronautical University (3)
- New Jersey Institute of Technology (3)
- Syracuse University (3)
- Technological University Dublin (3)
- University of Arkansas, Fayetteville (3)
- University of Kentucky (3)
- Ateneo de Manila University (2)
- CCT College Dublin (2)
- La Salle University (2)
- Ministry of Higher and Secondary Specialized Education of the Republic of Uzbekistan (2)
- Munster Technological University (2)
- Keyword
-
- Machine learning (12)
- Cybersecurity (9)
- Deep learning (8)
- Social media (7)
- Twitter (7)
-
- Big data (6)
- Information Technology (6)
- Internet of Things (6)
- Machine Learning (6)
- Cloud Computing (5)
- Cloud computing (5)
- Deep Learning (5)
- Information security (5)
- Clustering (4)
- Data mining (4)
- Deception (4)
- Fashion analysis (4)
- Fashion knowledge extraction (4)
- Natural Language Processing (4)
- Neural Networks (4)
- Security (4)
- Strategies (4)
- Visualization (4)
- Algorithm (3)
- Analytics (3)
- Artificial intelligence (3)
- Big Data (3)
- Center_CCR (3)
- Communication (3)
- Data (3)
- Publication
-
- Research Collection School Of Computing and Information Systems (140)
- Walden Dissertations and Doctoral Studies (53)
- Ryan Schuetzler (12)
- Master's Projects (11)
- Theses and Dissertations (7)
-
- Williams Honors College, Honors Research Projects (6)
- Electronic Theses and Dissertations (5)
- FIU Electronic Theses and Dissertations (5)
- Kno.e.sis Publications (5)
- Doctoral Dissertations (4)
- International Journal of Business and Technology (4)
- Masters Theses & Doctoral Dissertations (4)
- USF Tampa Graduate Theses and Dissertations (4)
- African Conference on Information Systems and Technology (3)
- Computer Science Faculty Publications (3)
- Honors Capstone Projects - All (3)
- Journal of Spatial Information Science (3)
- Bulletin of TUIT: Management and Communication Technologies (2)
- Computer Engineering (2)
- Conference papers (2)
- Copyright, Fair Use, Scholarly Communication, etc. (2)
- Department of Computer Science Publications (2)
- Department of Information Systems & Computer Science Faculty Publications (2)
- Dissertations (2)
- Dissertations and Theses Collection (Open Access) (2)
- Electrical and Computer Engineering Publications (2)
- Graduate Theses and Dissertations (2)
- ICT (2)
- Information Systems and Quantitative Analysis Faculty Publications (2)
- Master's Theses (2)
- Publication Type
Articles 1 - 30 of 358
Full-Text Articles in Physical Sciences and Mathematics
Early Detection Of Fake News On Social Media, Yang Liu
Early Detection Of Fake News On Social Media, Yang Liu
Dissertations
The ever-increasing popularity and convenience of social media enable the rapid widespread of fake news, which can cause a series of negative impacts both on individuals and society. Early detection of fake news is essential to minimize its social harm. Existing machine learning approaches are incapable of detecting a fake news story soon after it starts to spread, because they require certain amounts of data to reach decent effectiveness which take time to accumulate. To solve this problem, this research first analyzes and finds that, on social media, the user characteristics of fake news spreaders distribute significantly differently from those …
A Qualitative Representation Of Spatial Scenes In R2 With Regions And Lines, Joshua Lewis
A Qualitative Representation Of Spatial Scenes In R2 With Regions And Lines, Joshua Lewis
Electronic Theses and Dissertations
Regions and lines are common geographic abstractions for geographic objects. Collections of regions, lines, and other representations of spatial objects form a spatial scene, along with their relations. For instance, the states of Maine and New Hampshire can be represented by a pair of regions and related based on their topological properties. These two states are adjacent (i.e., they meet along their shared boundary), whereas Maine and Florida are not adjacent (i.e., they are disjoint).
A detailed model for qualitatively describing spatial scenes should capture the essential properties of a configuration such that a description of the represented objects …
Information Extraction From Biomedical Text Using Machine Learning, Deepti Garg
Information Extraction From Biomedical Text Using Machine Learning, Deepti Garg
Master's Projects
Inadequate drug experimental data and the use of unlicensed drugs may cause adverse drug reactions, especially in pediatric populations. Every year the U.S. Food and Drug Administration approves human prescription drugs for marketing. The labels associated with these drugs include information about clinical trials and drug response in pediatric population. In order for doctors to make an informed decision about the safety and effectiveness of these drugs for children, there is a need to analyze complex and often unstructured drug labels. In this work, first, an exploratory analysis of drug labels using a Natural Language Processing pipeline is performed. Second, …
Evaluating Data Quality Of Newborn Hearing Screening, Maria C. Sanchez Gomez, Kelly Dundon, Xidong Deng
Evaluating Data Quality Of Newborn Hearing Screening, Maria C. Sanchez Gomez, Kelly Dundon, Xidong Deng
Journal of Early Hearing Detection and Intervention
Scope
Jurisdictional-based Early Hearing Detection and Intervention Information Systems (EHDI-IS) collect data on the hearing screening and follow-up status of infants across the United States. These systems serve as tools that assist EHDI programs’ staff and partners in their tracking activities and provide a variety of data reports to help ensure that all children who are deaf/hard of hearing (D/HH) are identified early and receive recommended intervention services. The quality and timeliness of the data collected with these systems are crucial to effectively meeting these goals.
Methodology
Forty-eight EHDI programs, funded by the Centers for Disease Control and Prevention (CDC), …
Music Retrieval System Using Query-By-Humming, Parth Patel
Music Retrieval System Using Query-By-Humming, Parth Patel
Master's Projects
Music Information Retrieval (MIR) is a particular research area of great interest because there are various strategies to retrieve music. To retrieve music, it is important to find a similarity between the input query and the matching music. Several solutions have been proposed that are currently being used in the application domain(s) such as Query- by-Example (QBE) which takes a sample of an audio recording playing in the background and retrieves the result. However, there is no efficient approach to solve this problem in a Query-by-Humming (QBH) application. In a Query-by-Humming application, the aim is to retrieve music that is …
A Hybrid Approach For Multi-Document Text Summarization, Rashmi Varma
A Hybrid Approach For Multi-Document Text Summarization, Rashmi Varma
Master's Projects
Text summarization has been a long studied topic in the field of natural language processing. There have been various approaches for both extractive text summarization as well as abstractive text summarization. Summarizing texts for a single document is a methodical task. But summarizing multiple documents poses as a greater challenge. This thesis explores the application of Latent Semantic Analysis, Text-Rank, Lex-Rank and Reduction algorithms for single document text summarization and compares it with the proposed approach of creating a hybrid system combining each of the above algorithms, individually, with Restricted Boltzmann Machines for multi-document text summarization and analyzing how all …
Designing Learning Activities For Experiential Learning In A Design Thinking Course, Benjamin Gan, Eng Lieh Ouh
Designing Learning Activities For Experiential Learning In A Design Thinking Course, Benjamin Gan, Eng Lieh Ouh
Research Collection School Of Computing and Information Systems
One experiential learning design challenge is the duration of learning activities. These learning activities take up time and effort for teachers to design and student to perform. Another design challenge is the minimum instructional guidance of these learning activities which potentially impact the learning effectiveness of novice students. In this paper, we describe our findings of applying experiential learning method in a design thinking course with a list of learning activities performed iteratively. Each of the learning activity varies in their duration required and level of instructional guidance. Our survey seeks to find out which of the learning activities are …
Optimal Management Of Virtual Infrastructures Under Flexible Cloud Service Agreements, Zhiling Guo, Jin Li, Ram Ramesh
Optimal Management Of Virtual Infrastructures Under Flexible Cloud Service Agreements, Zhiling Guo, Jin Li, Ram Ramesh
Research Collection School Of Computing and Information Systems
A cloud service agreement entails the provisioning of a required set of virtual infrastructure resources at a specified level of availability to a client. The agreement also lays out the price charged to the client and a penalty to the provider when the assured availability is not met. The availability assurance involves backup resource provisioning, and the provider needs to allocate backups cost-effectively by balancing the resource-provisioning costs with the potential penalty costs. We develop stochastic dynamic optimization models of the backup resource-provisioning problem, leading to cost-effective resource-management policies in different practical settings. We present two sets of dynamic provisioning …
Hybrid Recommender Systems Via Spectral Learning And A Random Forest, Alyssa Williams
Hybrid Recommender Systems Via Spectral Learning And A Random Forest, Alyssa Williams
Electronic Theses and Dissertations
We demonstrate spectral learning can be combined with a random forest classifier to produce a hybrid recommender system capable of incorporating meta information. Spectral learning is supervised learning in which data is in the form of one or more networks. Responses are predicted from features obtained from the eigenvector decomposition of matrix representations of the networks. Spectral learning is based on the highest weight eigenvectors of natural Markov chain representations. A random forest is an ensemble technique for supervised learning whose internal predictive model can be interpreted as a nearest neighbor network. A hybrid recommender can be constructed by first …
Iomt Malware Detection Approaches: Analysis And Research Challenges, Mohammad Wazid, Ashok Kumar Das, Joel J.P.C. Rodrigues, Sachin Shetty, Youngho Park
Iomt Malware Detection Approaches: Analysis And Research Challenges, Mohammad Wazid, Ashok Kumar Das, Joel J.P.C. Rodrigues, Sachin Shetty, Youngho Park
VMASC Publications
The advancement in Information and Communications Technology (ICT) has changed the entire paradigm of computing. Because of such advancement, we have new types of computing and communication environments, for example, Internet of Things (IoT) that is a collection of smart IoT devices. The Internet of Medical Things (IoMT) is a specific type of IoT communication environment which deals with communication through the smart healthcare (medical) devices. Though IoT communication environment facilitates and supports our day-to-day activities, but at the same time it has also certain drawbacks as it suffers from several security and privacy issues, such as replay, man-in-the-middle, impersonation, …
A Transformative Concept: From Data Being Passive Objects To Data Being Active Subjects, Hans-Peter Plag, Shelley-Ann Jules-Plag
A Transformative Concept: From Data Being Passive Objects To Data Being Active Subjects, Hans-Peter Plag, Shelley-Ann Jules-Plag
OES Faculty Publications
The exploitation of potential societal benefits of Earth observations is hampered by users having to engage in often tedious processes to discover data and extract information and knowledge. A concept is introduced for a transition from the current perception of data as passive objects (DPO) to a new perception of data as active subjects (DAS). This transition would greatly increase data usage and exploitation, and support the extraction of knowledge from data products. Enabling the data subjects to actively reach out to potential users would revolutionize data dissemination and sharing and facilitate collaboration in user communities. The three core elements …
Strongly Leakage Resilient Authenticated Key Exchange, Revisited, Guomin Yang, Rongmao Chen, Yi Mu, Willy Susilo, Guo Fuchun, Jie Li
Strongly Leakage Resilient Authenticated Key Exchange, Revisited, Guomin Yang, Rongmao Chen, Yi Mu, Willy Susilo, Guo Fuchun, Jie Li
Research Collection School Of Computing and Information Systems
Authenticated Key Exchange (AKE) protocols allow two (or multiple) parties to authenticate each other and agree on a common secret key, which is essential for establishing a secure communication channel over a public network. AKE protocols form a central component in many network security standards such as IPSec, TLS/SSL, and SSH. However, it has been demonstrated that many standardized AKE protocols are vulnerable to side-channel and key leakage attacks. In order to defend against such attacks, leakage resilient (LR-) AKE protocols have been proposed in the literature. Nevertheless, most of the existing LR-AKE protocols only focused on the resistance to …
Quantum Consensus, Jorden Seet, Paul Griffin
Quantum Consensus, Jorden Seet, Paul Griffin
Research Collection School Of Computing and Information Systems
In this paper, we propose a novel consensus mechanism utilizing the quantum properties of qubits. This move from classical computing to quantum computing is shown to theoretically enhance the scalability and speed of distributed consensus as well as improve security and be a potential solution for the problem of blockchain interoperability. Using this method may circumvent the common problem known as the Blockchain Trilemma, enhancing scalability and speed without sacrificing de-centralization or byzantine fault tolerance. Consensus speed and scalability is shown by removing the need for multicast responses and exploiting quantum properties to ensure that only a single multicast is …
Objective Sleep Quality As A Predictor Of Mild Cognitive Impairment In Seniors Living Alone, Brian Chen, Hwee-Pink Tan, Irus Rawtaer, Hwee Xian Tan
Objective Sleep Quality As A Predictor Of Mild Cognitive Impairment In Seniors Living Alone, Brian Chen, Hwee-Pink Tan, Irus Rawtaer, Hwee Xian Tan
Research Collection School Of Computing and Information Systems
Singapore has the fastest ageing population in the Asia Pacific region, with an estimated 82,000 seniors living with dementia. These figures are projected to increase to more than 130,000 by 2030. The challenge is to identify more community dwelling seniors with Mild Cognitive Impairment (MCI), a prodromal state, as it provides an opportunity for evidence-based early intervention to delay the onset of dementia. In this paper, we explore the use of Internet of Things (IoT) systems in detecting MCI symptoms in seniors who are living alone, and accurately grouping them into MCI positive and negative subjects. We present feature extraction …
Influence, Information And Team Outcomes In Large Scale Software Development, Subhajit Datta
Influence, Information And Team Outcomes In Large Scale Software Development, Subhajit Datta
Research Collection School Of Computing and Information Systems
It is widely perceived that the egalitarian ecosystems of large scale open source software development foster effective team outcomes. In this study, we question this conventional wisdom by examining whether and how the centralization of information and influence in a software development team relate to the quality of the team's work products. Analyzing data from more than a hundred real world projects that include development activities over close to a decade, involving 2000+ developers, who collectively resolve more than two hundred thousand defects through discussions covering more than six hundred thousand comments, we arrive at statistically significant evidence indicating that …
Clustering Models For Topic Analysis In Graduate Discussion Forums, Mallika Gokarn Nitin, Swapna Gottipati, Venky Shankararaman
Clustering Models For Topic Analysis In Graduate Discussion Forums, Mallika Gokarn Nitin, Swapna Gottipati, Venky Shankararaman
Research Collection School Of Computing and Information Systems
Discussion forums provide the base content for creating a knowledge repository. It contains discussion threads related to key course topics that are debated by the students. In order to better understand the student learning experience, the instructor needs to analyse these discussion threads. This paper proposes the use of clustering models and interactive visualizations to conduct a qualitative analysis of graduate discussion forums. Our goal is to identify the sub-topics and topic evolutions in the discussion forums by applying text mining techniques. Our approach generates insights into the topic analysis in the forums and discovers the students’ cognitive understanding within …
The Dynamic Impacts Of Online Healthcare Community On Physician Altruism: A Hidden Markov Model, Kai Luo, Qiu-Hong Wang, Hock Hai Teo
The Dynamic Impacts Of Online Healthcare Community On Physician Altruism: A Hidden Markov Model, Kai Luo, Qiu-Hong Wang, Hock Hai Teo
Research Collection School Of Computing and Information Systems
Physician altruism is not only a key foundation of modern medical professionalism, but also a critical component in the theoretical health economics study. There is considerable interest in understanding the impacts of contemporary healthcare technology on physician altruism. In this paper, we investigate the dynamic influence of multiple incentive mechanisms developed by an online healthcare community (OHC) on physician altruism. We model physician altruism as the degree of tendency to benefit the patients at the cost of oneself and focus on the incentive mechanisms that give physicians social and economic returns. The dynamics of physician altruism is characterized via a …
Examining The Theoretical Mechanisms Underlying Health Information Exchange Impact On Healthcare Outcomes: A Physician Agency Perspective, Fang Zhou, Qiu-Hong Wang, Hock Hai Teo
Examining The Theoretical Mechanisms Underlying Health Information Exchange Impact On Healthcare Outcomes: A Physician Agency Perspective, Fang Zhou, Qiu-Hong Wang, Hock Hai Teo
Research Collection School Of Computing and Information Systems
Health information exchange (HIE) is presumed to reduce medical costs by facilitating information sharing across healthcare providers. Existing studies focused on different medical costs or one set of costs, and resulted in mixed findings. We examine the effects of patient access to HIE on two of the most important medical costs of a hospitalization episode - test costs and medication costs - through a natural experiment and the discharge data of a hospital. Besides the negative direct effect of access to HIT on tests costs, we also find its positive spillover effect on medication costs, such that more patients having …
Happy Toilet: A Social Analytics Approach To The Study Of Public Toilet Cleanliness, Eugene W. J. Choy, Winston M. K. Ho, Xiaohang Li, Ragini Verma, Li Jin Sim, Kyong Jin Shim
Happy Toilet: A Social Analytics Approach To The Study Of Public Toilet Cleanliness, Eugene W. J. Choy, Winston M. K. Ho, Xiaohang Li, Ragini Verma, Li Jin Sim, Kyong Jin Shim
Research Collection School Of Computing and Information Systems
This study presents a social analytics approach to the study of public toilet cleanliness in Singapore. From popular social media platforms, our system automatically gathers and analyzes relevant public posts that mention about toilet cleanliness in highly frequented locations across the Singapore island - from busy shopping malls to food 'hawker' centers.
Optimal Design And Ownership Structures Of Innovative Retail Payment Systems, Zhiling Guo, Dan Ma
Optimal Design And Ownership Structures Of Innovative Retail Payment Systems, Zhiling Guo, Dan Ma
Research Collection School Of Computing and Information Systems
In response to the Fintech trend, an ongoing debate in the banking industry is how to design the new-generation interbank retail payment and settlement system. We propose a two-stage analytical model that takes into account the value-risk tradeoff in the new payment system design, as well as banks’ participation incentives and adoption timing decisions. We find that, as the system base value increases, banks tend to synchronize their investment and adoption decisions. When the system base value is low and banks are heterogeneous, bank association ownership maximizes social welfare. When both the system base value and bank heterogeneity are moderate, …
Gms: Grid-Based Motion Statistics For Fast, Ultra-Robust Feature Correspondence, Jia-Wang Bian, Wen-Yan Lin, Yun Liu, Le Zhang, Sai-Kit Yeung, Ming-Ming Cheng, Ian Reid
Gms: Grid-Based Motion Statistics For Fast, Ultra-Robust Feature Correspondence, Jia-Wang Bian, Wen-Yan Lin, Yun Liu, Le Zhang, Sai-Kit Yeung, Ming-Ming Cheng, Ian Reid
Research Collection School Of Computing and Information Systems
Feature matching aims at generating correspondences across images, which is widely used in many computer vision tasks. Although considerable progress has been made on feature descriptors and fast matching for initial correspondence hypotheses, selecting good ones from them is still challenging and critical to the overall performance. More importantly, existing methods often take a long computational time, limiting their use in real-time applications. This paper attempts to separate true correspondences from false ones at high speed. We term the proposed method (GMS) grid-based motion Statistics, which incorporates the smoothness constraint into a statistic framework for separation and uses a grid-based …
Improved Generalisation Bounds For Deep Learning Through L∞ Covering Numbers, Antoine Ledent, Yunwen Lei, Marius Kloft
Improved Generalisation Bounds For Deep Learning Through L∞ Covering Numbers, Antoine Ledent, Yunwen Lei, Marius Kloft
Research Collection School Of Computing and Information Systems
Using proof techniques involving L∞ covering numbers, we show generalisation error bounds for deep learning with two main improvements over the state of the art. First, our bounds have no explicit dependence on the number of classes except for logarithmic factors. This holds even when formulating the bounds in terms of the L 2 norm of the weight matrices, while previous bounds exhibit at least a square-root dependence on the number of classes in this case. Second, we adapt the Rademacher analysis of DNNs to incorporate weight sharing—a task of fundamental theoretical importance which was previously attempted only under very …
Sentiment Analysis, Quantification, And Shift Detection, Kevin Labille
Sentiment Analysis, Quantification, And Shift Detection, Kevin Labille
Graduate Theses and Dissertations
This dissertation focuses on event detection within streams of Tweets based on sentiment quantification. Sentiment quantification extends sentiment analysis, the analysis of the sentiment of individual documents, to analyze the sentiment of an aggregated collection of documents. Although the former has been widely researched, the latter has drawn less attention but offers greater potential to enhance current business intelligence systems. Indeed, knowing the proportion of positive and negative Tweets is much more valuable than knowing which individual Tweets are positive or negative. We also extend our sentiment quantification research to analyze the evolution of sentiment over time to automatically detect …
Extracting Social Network From Literary Prose, Tarana Tasmin Bipasha
Extracting Social Network From Literary Prose, Tarana Tasmin Bipasha
Graduate Theses and Dissertations
This thesis develops an approach to extract social networks from literary prose, namely, Jane Austen’s published novels from eighteenth- and nineteenth- century. Dialogue interaction plays a key role while we derive the networks, thus our technique relies upon our ability to determine when two characters are in conversation. Our process involves encoding plain literary text into the Text Encoding Initiative’s (TEI) XML format, character name identification, conversation and co-occurrence detection, and social network construction. Previous work in social network construction for literature have focused on drama, specifically manually TEI-encoded Shakespearean plays in which character interactions are much easier to track …
Study Group Travel Behaviour Patterns From Large-Scale Smart Card Data, Xiancai Tian, Baihua Zheng
Study Group Travel Behaviour Patterns From Large-Scale Smart Card Data, Xiancai Tian, Baihua Zheng
Research Collection School Of Computing and Information Systems
In this paper, we aim at studying the group travel behaviour (GTB) patterns from large-scale auto fare collection (AFC) data. GTB is defined as two or more commuters intentionally and regularly traveling together from an origin to a destination. We propose a method to identify GTB accurately and efficiently and apply our method to the Singapore AFC dataset to reveal the GTB patterns of Singapore commuters. The case study proves that our method is able to identify GTB patterns more accurately and efficiently than the state-of-the-art.
Self-Organizing Neural Networks For Universal Learning And Multimodal Memory Encoding, Ah-Hwee Tan, Budhitama Subagdja, Di Wang, Lei Meng
Self-Organizing Neural Networks For Universal Learning And Multimodal Memory Encoding, Ah-Hwee Tan, Budhitama Subagdja, Di Wang, Lei Meng
Research Collection School Of Computing and Information Systems
Learning and memory are two intertwined cognitive functions of the human brain. This paper shows how a family of biologically-inspired self-organizing neural networks, known as fusion Adaptive Resonance Theory (fusion ART), may provide a viable approach to realizing the learning and memory functions. Fusion ART extends the single-channel Adaptive Resonance Theory (ART) model to learn multimodal pattern associative mappings. As a natural extension of ART, various forms of fusion ART have been developed for a myriad of learning paradigms, ranging from unsupervised learning to supervised learning, semi-supervised learning, multimodal learning, reinforcement learning, and sequence learning. In addition, fusion ART models …
Salience-Aware Adaptive Resonance Theory For Large-Scale Sparse Data Clustering, Lei Meng, Ah-Hwee Tan, Chunyan Miao
Salience-Aware Adaptive Resonance Theory For Large-Scale Sparse Data Clustering, Lei Meng, Ah-Hwee Tan, Chunyan Miao
Research Collection School Of Computing and Information Systems
Sparse data is known to pose challenges to cluster analysis, as the similarity between data tends to be ill-posed in the high-dimensional Hilbert space. Solutions in the literature typically extend either k-means or spectral clustering with additional steps on representation learning and/or feature weighting. However, adding these usually introduces new parameters and increases computational cost, thus inevitably lowering the robustness of these algorithms when handling massive ill-represented data. To alleviate these issues, this paper presents a class of self-organizing neural networks, called the salience-aware adaptive resonance theory (SA-ART) model. SA-ART extends Fuzzy ART with measures for cluster-wise salient feature modeling. …
Identifying Regional Trends In Avatar Customization, Peter Mawhorter, Sercan Sengun, Haewoon Kwak, D. Fox Harrell
Identifying Regional Trends In Avatar Customization, Peter Mawhorter, Sercan Sengun, Haewoon Kwak, D. Fox Harrell
Research Collection School Of Computing and Information Systems
Since virtual identities such as social media profiles and avatars have become a common venue for self-expression, it has become important to consider the ways in which existing systems embed the values of their designers. In order to design virtual identity systems that reflect the needs and preferences of diverse users, understanding how the virtual identity construction differs between groups is important. This paper presents a new methodology that leverages deep learning and differential clustering for comparative analysis of profile images, with a case study of almost 100 000 avatars from a large online community using a popular avatar creation …
A Unified Variance-Reduced Accelerated Gradient Method For Convex Optimization, Guanghui Lan, Zhize Li, Yi Zhou
A Unified Variance-Reduced Accelerated Gradient Method For Convex Optimization, Guanghui Lan, Zhize Li, Yi Zhou
Research Collection School Of Computing and Information Systems
We propose a novel randomized incremental gradient algorithm, namely, VAriance-Reduced Accelerated Gradient (Varag), for finite-sum optimization. Equipped with a unified step-size policy that adjusts itself to the value of the conditional number, Varag exhibits the unified optimal rates of convergence for solving smooth convex finite-sum problems directly regardless of their strong convexity. Moreover, Varag is the first accelerated randomized incremental gradient method that benefits from the strong convexity of the data-fidelity term to achieve the optimal linear convergence. It also establishes an optimal linear rate of convergence for solving a wide class of problems only satisfying a certain error bound …
Ssrgd: Simple Stochastic Recursive Gradient Descent For Escaping Saddle Points, Zhize Li
Ssrgd: Simple Stochastic Recursive Gradient Descent For Escaping Saddle Points, Zhize Li
Research Collection School Of Computing and Information Systems
We analyze stochastic gradient algorithms for optimizing nonconvex problems. In particular, our goal is to find local minima (second-order stationary points) instead of just finding first-order stationary points which may be some bad unstable saddle points. We show that a simple perturbed version of stochastic recursive gradient descent algorithm (called SSRGD) can find an $(\epsilon,\delta)$-second-order stationary point with $\widetilde{O}(\sqrt{n}/\epsilon^2 + \sqrt{n}/\delta^4 + n/\delta^3)$ stochastic gradient complexity for nonconvex finite-sum problems. As a by-product, SSRGD finds an $\epsilon$-first-order stationary point with $O(n+\sqrt{n}/\epsilon^2)$ stochastic gradients. These results are almost optimal since Fang et al. [2018] provided a lower bound $\Omega(\sqrt{n}/\epsilon^2)$ for finding …