Computer Sciences Commons^™

Seapot-Rl: Selective Exploration Algorithm For Policy Transfer In Rl, Akshay Narayan, Zhuoru Li, Tze-Yun Leong

Research Collection School Of Computing and Information Systems

We propose a new method for transferring a policy from a source task to a target task in model-based reinforcement learning. Our work is motivated by scenarios where a robotic agent operates in similar but challenging environments, such as hospital wards, differentiated by structural arrangements or obstacles, such as furniture. We address problems that require fast responses adapted from incomplete, prior knowledge of the agent in new scenarios. We present an efficient selective exploration strategy that maximally reuses the source task policy. Reuse efficiency is effected through identifying sub-spaces that are different in the target environment, thus limiting the exploration …

Unsupervised Visual Hashing With Semantic Assistant For Content-Based Image Retrieval, Lei Zhu, Jialie Shen, Liang Xie, Zhiyong Cheng

Research Collection School Of Computing and Information Systems

As an emerging technology to support scalable content-based image retrieval (CBIR), hashing has recently received great attention and became a very active research domain. In this study, we propose a novel unsupervised visual hashing approach called semantic-assisted visual hashing (SAVH). Distinguished from semi-supervised and supervised visual hashing, its core idea is to effectively extract the rich semantics latently embedded in auxiliary texts of images to boost the effectiveness of visual hashing without any explicit semantic labels. To achieve the target, a unified unsupervised framework is developed to learn hash codes by simultaneously preserving visual similarities of images, integrating the semantic …

An Efficient Approach To Model-Based Hierarchical Reinforcement Learning, Zhuoru Li, Akshay Narayan, Tze-Yun Leong

Research Collection School Of Computing and Information Systems

We propose a model-based approach to hierarchical reinforcement learning that exploits shared knowledge and selective execution at different levels of abstraction, to efficiently solve large, complex problems. Our framework adopts a new transition dynamics learning algorithm that identifies the common action-feature combinations of the subtasks, and evaluates the subtask execution choices through simulation. The framework is sample efficient, and tolerates uncertain and incomplete problem characterization of the subtasks. We test the framework on common benchmark problems and complex simulated robotic environments. It compares favorably against the stateof-the-art algorithms, and scales well in very large problems.

Active Video Summarization: Customized Summaries Via On-Line Interaction With The User, Ana Garcia Del Molino, Xavier Boix, Joo-Hwee Lim, Ah-Hwee Tan

Research Collection School Of Computing and Information Systems

To facilitate the browsing of long videos, automatic video summarization provides an excerpt that represents its content. In the case of egocentric and consumer videos, due to their personal nature, adapting the summary to specific user’s preferences is desirable. Current approaches to customizable video summarization obtain the user’s preferences prior to the summarization process. As a result, the user needs to manually modify the summary to further meet the preferences. In this paper, we introduce Active Video Summarization (AVS), an interactive approach to gather the user’s preferences while creating the summary. AVS asks questions about the summary to update it …

Optimizing Expectation With Guarantees In Pomdps, Krishnendu Chatterjee, Guillermo A. Perez, Jean-François Raskin, Dorde Zikelic

Research Collection School Of Computing and Information Systems

A standard objective in partially-observable Markov decision processes (POMDPs) is to find a policy that maximizes the expected discounted-sum payoff. However, such policies may still permit unlikely but highly undesirable outcomes, which is problematic especially in safety-critical applications. Recently, there has been a surge of interest in POMDPs where the goal is to maximize the probability to ensure that the payoff is at least a given threshold, but these approaches do not consider any optimization beyond satisfying this threshold constraint. In this work we go beyond both the "expectation" and "threshold" approaches and consider a "guaranteed payoff optimization (GPO)" problem …

Using An Online Tutorial To Teach Rea Data Modeling In Accounting Information Systems Courses, Poh Sun Seow, Pan, Gary

Research Collection School Of Accountancy

Online learning has been gaining widespread adoption due to its successin enhancing student-learning outcomes and improving student t academicperformance. This paper describes an online tutorial to teach resource-event-agent(REA) data modeling in an undergraduate accounting information systems course.The REA online tutorial reflects a self-study application designed to helpstudents improve their understanding of the REA data model. As such, thetutorial acts as a supplement to lectures by reinforcing the concepts andincorporating practices to assess student understanding. Instructors can accessthe REA online tutorial at http://smu.asg/rea. An independent survey by the University’sCentre for Teaching Excellence found a significant increase in students’perceived knowledge of REA …

Bike Route Choice Modeling Using Gps Data Without Choice Sets Of Paths, Maëlle Zimmermann, Tien Mai, Emma Frejinger

Research Collection School Of Computing and Information Systems

Concerned by the nuisances of motorized travel on urban life, policy makers are faced with the challenge of making cycling a more attractive alternative for everyday transportation. Route choice models can help achieve this objective by gaining insights into the trade-offs cyclists make when choosing their routes and by allowing the effect of infrastructure improvements to be analyzed. We estimate a link-based bike route choice model from a sample of GPS observations in the city of Eugene on a network comprising over 40,000 links. The so-called recursive logit (RL) model (Fosgerau et al., 2013) does not require to sample any …

Helping Businesses Embrace Transformational Technology, Singapore Management University

Research@SMU: Connecting the Dots

Research from the SMU School of Information Systems is helping institutions understand both the benefits and risks of real-time business transactions.

Professor Ma, together with SIS colleagues Professors Robert Kauffman, Guo Zhiling and Mei Lin, are investigators on a project studying the feasibility of implementing real-time payment and settlement systems.

See the paper:

• Near real-time retail payment and settlement systems mechanism design
• Mechanism design for near real-time retail payment and settlement systems

Harnessing Technology To Improve The Consumer Experience, Singapore Management University

Research@SMU: Connecting the Dots

Professor Guo Zhiling is studying the impact of information technology on consumer behaviour and business strategy.

See the papers:

• Assessing the impact of recommendation agents on on-line consumer unplanned purchase behavior
• Assessing the moderating effect of consumer product knowledge and online shopping experience on using recommendation agents for customer loyalty

Walking Firms Through It Decisions, Singapore Management University

Research@SMU: Connecting the Dots

Is it better to be an early adopter or wait till the technology matures? Professor Ma Dan helps companies make data-driven decisions.

See the papers:

• The pricing strategy analysis for the Software-as-a-Service business model
• Cloud services from a consumer perspective.

Using Smartphones To Turn Cities Into Innovation Engines, Singapore Management University

Research@SMU: Connecting the Dots

Can smartphones improve our quality of life and make cities run more efficiently? Professor Rajesh Balan is working towards this vision.

See the papers:

• Real-time trip information service for a large taxi fleet
• Smartwatch based food diary and eating analytics
• Need accurate user behaviour?: Pay attention to groups!

Pushing Boundaries In Software Analytics, Singapore Management University

Research@SMU: Connecting the Dots

Professor David Lo is working to push the boundaries of what can be done by analysing huge amounts of software data.

See the papers:

• AutoQuery: Automatic construction of dependency queries for code search
• Information retrieval and spectrum based bug localization: Better together

Keeping Data Safe From Prying Eyes, Singapore Management University

Research@SMU: Connecting the Dots

An expert in information security research, Professor Robert Deng is constantly trying to beat hackers at their own game.

See the papers:

• Launching generic attacks on iOS with approved third-party applications
• Universally composable RFID mutual authentication
• Attribute-based cloud storage with secure provenance over encrypted data

Smart Homes Enhance Seniors’ Safety, Singapore Management University

Research@SMU: Connecting the Dots

Professor Tan Hwee Pink and researchers at iCity Lab are using sensors to increase the safety of seniors who live independently in their own homes.

See the papers:

• Online detection of behavioral change using unobtrusive eldercare monitoring system
• Improving the sensitivity of unobtrusive inactivity detection in sensor-enabled homes for the elderly
• SHINESeniors: Personalized services for active ageing-in-place

Strengthening Cybersecurity Through Research, Singapore Management University

Research@SMU: Connecting the Dots

Tackling cybersecurity risks in mobile computing by Professors Robert Deng and Li Yingjiu.

See their papers:

• Seeing your face is not enough: An inertial sensor-based liveness detection for face authentication
• An efficient privacy-preserving outsourced computation over public data
• Secure encrypted data deduplication with ownership proof and user revocation

Managing Cities With Urban Computing, Singapore Management University

Research@SMU: Connecting the Dots

SMU Professors Lau Hoong Chuin and Lim Yun Fong are combining their mathematical, computational and business know-how to address challenges facing the inner-city deliveries of orders and shipments.

See the papers:

• A rolling horizon auction mechanism and virtual pricing of shipping capacity for urban consolidation centers
• Achieving economic and environmental sustainabilities in urban consolidation center with bicriteria auction
• Retail precinct management: A case of commercial decentralization in Singapore

Research@Smu: Connecting The Dots, Singapore Management University

Research@SMU: Connecting the Dots

The word “Management” in the name of our university was carefully and deliberately chosen to address the needs of institutions and decisionmakers across all segments of society: business enterprises and the private sector, government and the public sector, and civil society and the people sector. Unless you are already familiar with the breadth and depth of research output from Singapore Management University (SMU), you would naturally assume that we were a “management” university that was focusing exclusively on business – in essence, a large business school. The reality is quite different.

Through this book, you will gain a more complete …

Using An Online Learning Tutorial To Teach Rea Data Modelling In Accounting Information Systems Courses, Poh Sun Seow, Gary Pan

Research Collection School Of Accountancy

Online learning has been gaining widespread adoption due to its success in enhancing student-learning outcomes and improving student academic performance. This paper describes an online tutorial to teach resource-event-agent (REA) data modeling in an undergraduate accounting information systems course. The REA online tutorial reflects a self-study application designed to help students improve their understanding of the REA data model. As such, the tutorial acts as a supplement to lectures by reinforcing the concepts and incorporating practices to assess student understanding. Instructors can access the REA online tutorial at http://smu.sg/rea. An independent survey by the University's Centre for Teaching Excellence found …

The Habits Of Highly Effective Researchers: An Empirical Study, Subhajit Datta, Partha Basuchowdhuri, Surajit Acharya, Subhashis Majumder

Research Collection School Of Computing and Information Systems

Interest in the habits of influential individuals cuts across domains. As researchers, we are intrigued why few attain significant eminence in their fields, whereas many operate in obscurity. An empirical examination of this question has been made possible by the recent availability of large scale publication data. In this paper, we use information from the AMiner Paper Citation and Author Collaboration Networks to discern factors that relate to the impact of influential researchers across five domains in the computing discipline. We propose and apply a novel algorithm to identify influential vertices in co-authorship networks built from total corpora of 1,00,000+papers …

A Two-Sided Collaborative Transparent Display Supporting Workspace Awareness, Jiannan Li, Saul Greenberg, Ehud Sharlin

Research Collection School Of Computing and Information Systems

Transparent displays naturally support workspace awareness during face-to-face interactions. Viewers see another person’s actions through the display: gestures, gaze, body movements, and what one is manipulating on the display. Yet we can design even better collaborative transparent displays. First, collaborators on either side should be able to directly interact with workspace objects. Second, and more controversially, both sides should be capable of presenting different content. This affords: reversal of images/text in place (so that people on both sides see objects correctly); personal and private territories aligned atop each other; and GUI objects that provide different visuals for feedthrough vs. feedback. …

Concept-Based Interactive Search System, Yi-Jie Lu, Phuong Anh Nguyen, Hao Zhang, Chong-Wah Ngo

Research Collection School Of Computing and Information Systems

Our successful multimedia event detection system at TREC-VID 2015 showed its strength on handling complex concepts in a query. The system was based on a large number of pre-trained concept detectors for textual-to-visual relation. In this paper, we enhance the system by enabling human-in-the-loop. In order to facilitate a user to quickly find an information need, we incorporate concept screening, video reranking by highlighted concepts, relevance feedback and color sketch to refine a coarse retrieval result. The aim is to eventually come up with a system suitable for both Ad-hoc Video Search and Known-Item Search. In addition, as the increasing …

Cross-Modal Recipe Retrieval: How To Cook This Dish?, Jingjing Chen, Lei Pang, Chong-Wah Ngo

Research Collection School Of Computing and Information Systems

In social media users like to share food pictures. One intelligent feature, potentially attractive to amateur chefs, is the recommendation of recipe along with food. Having this feature, unfortunately, is still technically challenging. First, the current technology in food recognition can only scale up to few hundreds of categories, which are yet to be practical for recognizing ten of thousands of food categories. Second, even one food category can have variants of recipes that differ in ingredient composition. Finding the best-match recipe requires knowledge of ingredients, which is a fine-grained recognition problem. In this paper, we consider the problem from …

Enabling Sustainable Bulk Transfer In Environmentally-Powered Wireless Sensor Networks, Alvin C. Valera, Wee-Seng Soh, Hwee-Pink Tan

Research Collection School Of Computing and Information Systems

We address the problem of transferring bulk data in environmentally-powered wireless sensor networks where duty cycle compliance is critical for their uninterrupted operation. We propose Pump-and-Nap, a packet train forwarding technique that maximizes throughput while simultaneously enforcing compliance to dynamic duty cycle limitations. A node using Pump-and-Nap operates by pumping a train of packets followed by a napping period where the node forgoes any transmission. Pump-and-Nap employs an adaptive controller to periodically compute the optimal capacity, that is, the maximum number of packets a node can receive and transmit in a train, given its duty cycle constraint. The controller uses …

Universally Composable Rfid Mutual Authentication, Chunhua Su, Bagus Santoso, Yingjiu Li, Robert H. Deng, Xinyi Huang

Research Collection School Of Computing and Information Systems

Universally Composable (UC) framework provides the strongest security notion for designing fully trusted cryptographic protocols, and it is very challenging on applying UC security in the design of RFID mutual authentication protocols. In this paper, we formulate the necessary conditions for achieving UC secure RFID mutual authentication protocols which can be fully trusted in arbitrary environment, and indicate the inadequacy of some existing schemes under the UC framework. We define the ideal functionality for RFID mutual authentication and propose the first UC secure RFID mutual authentication protocol based on public key encryption and certain trusted third parties which can be …

Online Hacker Forum Censorship: Would Banning The Bad Guys Attract Good Guys?, Qiu-Hong Wang, Le-Ting Zhang, Meng-Ke Qiao

Research Collection School Of Computing and Information Systems

To tackle the ubiquitous cybersecurity threats, a few countries have enacted legislation to criminalize the production, distribution and possession of computer misuse tools. Consequently, online hacker forums, which enable the provision and dissemination of malicious cyber-attack techniques among potential hackers or technology-savvy users, are subject to censorship. This project examines the mixed impacts of online hacker forum censorship on users’ contribution to protection discussion through a natural experiment with large-scale content analysis. We find that while the enforcement indeed reduced the discussion on malicious cyber-attacks, the discussion on cybersecurity protection could increase or decrease in different scenarios. The rationale is …

On The Effectiveness Of Code-Reuse-Based Android Application Obfuscation, Xiaoxiao Tang, Yu Liang, Xinjie Ma, Yan Lin, Debin Gao

Research Collection School Of Computing and Information Systems

Attackers use reverse engineering techniques to gain detailed understanding of executable for malicious purposes, such as re-packaging an Android app to inject malicious code or advertising components. To make reverse engineering more difficult, researchers have proposed various code obfuscation techniques to conceal purposes or logic of code segments. One interesting idea of code obfuscation is to apply codereuse techniques (e.g., Return-Oriented Programming) to (re-)distribute essential code segments before they are reconstructed at runtime. Such techniques are well understood on x86 platform, but relatively less explored on Android. In this paper, we present an evaluation on the extent to which code-reuse-based …

Discovering Historic Traffic-Tolerant Paths In Road Networks, Pui Hang Li, Man Lung Yiu, Kyriakos Mouratidis

Research Collection School Of Computing and Information Systems

Historic traffic information is valuable in transportation analysis and planning, e.g., evaluating the reliability of routes for representative source-destination pairs. Also, it can be utilized to provide efficient and effective route-search services. In view of these applications, we propose the k traffic-tolerant paths (TTP) problem on road networks, which takes a source-destination pair and historic traffic information as input, and returns k paths that minimize the aggregate (historic) travel time. Unlike the shortest path problem, the TTP problem has a combinatorial search space that renders the optimal solution expensive to find. First, we propose an exact algorithm with effective pruning …

Easing Cross-Border Communication: Mobile-Mediated Communication And Its Framework, Kyungsub Choi, Young Soo Kim

Research Collection School Of Computing and Information Systems

Communication is probably the most critical component of an organization engaged in a cross-border collaboration. Today’s smart devices substantially contribute to such communication. Combined with social media, mobile communication technologies are becoming the main platform for many core functions within organizations. In this paper, we identified seven media identifiable attributes: synchronicity (SYN), de-individuation and co-presence (DCP), accessibility readiness (ARD), cognizance of environment change (CEC), wearability-portability (WRB) modality-select (MDS) and visibility (VSB). These seven attributes significantly impact the course of mobile-mediated communication. We believe that development of a theoretical perspective that embraces the complexity of mobile-mediated communication is due in order …

Technology Mergers And Acquisitions In The Presence Of An Installed Base: A Strategic Analysis, Qiu-Hong Wang, Kai-Lung Hui

Research Collection School Of Computing and Information Systems

We study the strategic benefits of M&A when competing IT vendors sell different generations of the same product with different quality. We assume the new product arrives unexpectedly when an installed base of old product exists. We show that the combination of consumers' purchase history and heterogeneity leads to new demand complexity that gives rise to innovative product strategies. We find that shelving the old product is an important motivation for M&A. The acquirer may exercise static or intertemporal price discrimination depending on whether it can exercise upgrade pricing. M&A may speed up or slow down new product consumption, and …

Exploring Representativeness And Informativeness For Active Learning, Bo Du, Zengmao Wang, Lefei Zhang, Liangpei Zhang, Wei Liu, Jialie Shen, Dacheng Tao

Research Collection School Of Computing and Information Systems

How can we find a general way to choose the most suitable samples for training a classifier? Even with very limited prior information? Active learning, which can be regarded as an iterative optimization procedure, plays a key role to construct a refined training set to improve the classification performance in a variety of applications, such as text analysis, image recognition, social network modeling, etc. Although combining representativeness and informativeness of samples has been proven promising for active sampling, state-of-the-art methods perform well under certain data structures. Then can we find a way to fuse the two active sampling criteria without …