Computer Sciences Commons^™

Mlp-3d: A Mlp-Like 3d Architecture With Grouped Time Mixing, Zhaofan Qiu, Ting Yao, Chong-Wah Ngo, Tao Mei

Research Collection School Of Computing and Information Systems

Convolutional Neural Networks (CNNs) have been re-garded as the go-to models for visual recognition. More re-cently, convolution-free networks, based on multi-head self-attention (MSA) or multi-layer perceptrons (MLPs), become more and more popular. Nevertheless, it is not trivial when utilizing these newly-minted networks for video recognition due to the large variations and complexities in video data. In this paper, we present MLP-3D networks, a novel MLP-like 3D architecture for video recognition. Specifically, the architecture consists of MLP-3D blocks, where each block contains one MLP applied across tokens (i.e., token-mixing MLP) and one MLP applied independently to each token (i.e., channel MLP). …

Class Re-Activation Maps For Weakly-Supervised Semantic Segmentation, Zhaozheng Chen, Tan Wang, Xiongwei Wu, Xian-Sheng Hua, Hanwang Zhang, Qianru Sun

Research Collection School Of Computing and Information Systems

Extracting class activation maps (CAM) is arguably the most standard step of generating pseudo masks for weakly supervised semantic segmentation (WSSS). Yet, we find that the crux of the unsatisfactory pseudo masks is the binary cross-entropy loss (BCE) widely used in CAM. Specifically, due to the sum-over-class pooling nature of BCE, each pixel in CAM may be responsive to multiple classes co-occurring in the same receptive field. To this end, we introduce an embarrassingly simple yet surprisingly effective method: Reactivating the converged CAM with BCE by using softmax crossentropy loss (SCE), dubbed ReCAM. Given an image, we use CAM to …

Revisiting Local Descriptor For Improved Few-Shot Classification, Jun He, Richang Hong, Xueliang Liu, Mingliang Xu, Qianru Sun

Research Collection School Of Computing and Information Systems

Few-shot classification studies the problem of quickly adapting a deep learner to understanding novel classes based on few support images. In this context, recent research efforts have been aimed at designing more and more complex classifiers that measure similarities between query and support images but left the importance of feature embeddings seldom explored. We show that the reliance on sophisticated classifiers is not necessary, and a simple classifier applied directly to improved feature embeddings can instead outperform most of the leading methods in the literature. To this end, we present a new method, named DCAP, for few-shot classification, in which …

Adaptive Temporal Grouping For Black-Box Adversarial Attacks On Videos, Zhipeng Wei, Jingjing Chen, Hao Zhang, Linxi Jiang, Yu-Gang Jiang

Research Collection School Of Computing and Information Systems

Deep-learning based video models, which have remarkable performance on action recognition tasks, are recently proved to be vulnerable to adversarial samples, even those generated in the black-box setting. However, these black-box attack methods are insufficient to attack videos models in real-world applications due to the requirement of lots of queries. To this end, we propose to boost the efficiency of black-box attacks on video recognition models. Although videos carry rich temporal information, they include redundant spatial information from adjacent frames. This motivates us to introduce the adaptive temporal grouping (ATG) method, which groups video frames by the similarity of their …

Communicative Strategies For Building Public Confidence In Data Governance: Analyzing Singapore's Covid-19 Contact-Tracing Initiatives, Gordon Kuo Siong Tan, Sun Sun Lim

Research Collection College of Integrative Studies

Effective social data governance rests on a bedrock of social support. Without securing trust from the populace whose information is being collected, analyzed, and deployed, policies on which such data are based will be undermined by a lack of public confidence. The COVID-19 pandemic has accelerated digitalization and datafication by governments for the purposes of contact tracing and epidemiological investigation. However, concerns about surveillance and data privacy have stunted the adoption of such contact-tracing initiatives. This commentary analyzes Singapore's contact-tracing initiative to uncover the reasons for public resistance and efforts by the state to address them. The government's contact-tracing program …

Learnings From A Pilot Hybrid Question Answering System: Cqas: Case Study Based On A Singapore Government Agency's Customer Service Centre, Hui Shan Lee, Shankararaman, Venky, Eng Lieh Ouh

Research Collection School Of Computing and Information Systems

The Singapore Government first released their digital government blueprint in 2018 with the key message for all their agencies to be "digital to the core and served with heart". With this push, agencies are moving towards human-centric digital services, especially for individual citizens. During COVID-19, Singapore government agencies introduced many COVID-19 digital initiatives resulting in more incoming inquiries from citizens to respective agencies. This surge in inquiries created the challenge on the agencies' end to meet service level agreements. One widely adopted solution is the use of chatbot technology that directly interfaces with the customer. However, several organisations have faced …

Generative Flows With Invertible Attentions, Rhea Sanjay Sukthanker, Zhiwu Huang, Suryansh Kumar, Radu Timofte, Luc Van Gool

Research Collection School Of Computing and Information Systems

Flow-based generative models have shown an excellent ability to explicitly learn the probability density function of data via a sequence of invertible transformations. Yet, learning attentions in generative flows remains understudied, while it has made breakthroughs in other domains. To fill the gap, this paper introduces two types of invertible attention mechanisms, i.e., map-based and transformer-based attentions, for both unconditional and conditional generative flows. The key idea is to exploit a masked scheme of these two attentions to learn long-range data dependencies in the context of generative flows. The masked scheme allows for invertible attention modules with tractable Jacobian determinants, …

Officers: Operational Framework For Intelligent Crime-And-Emergency Response Scheduling, Jonathan David Chase, Siong Thye Goh, Tran Phong, Hoong Chuin Lau

Research Collection School Of Computing and Information Systems

In the quest to achieve better response times in dense urban environments, law enforcement agencies are seeking AI-driven planning systems to inform their patrol strategies. In this paper, we present a framework, OFFICERS, for deployment planning that learns from historical data to generate deployment schedules on a daily basis. We accurately predict incidents using ST-ResNet, a deep learning technique that captures wide-ranging spatio-temporal dependencies, and solve a large-scale optimization problem to schedule deployment, significantly improving its scalability through a simulated annealing solver. Methodologically, our approach outperforms our previous works where prediction was done using Generative Adversarial Networks, and optimization was …

Joint Pricing And Matching For City-Scale Ride Pooling, Sanket Shah, Meghna Lowalekar, Pradeep Varakantham

Research Collection School Of Computing and Information Systems

Central to efficient ride-pooling are two challenges: (1) how to `price' customers' requests for rides, and (2) if the customer agrees to that price, how to best `match' these requests to drivers. While both of them are interdependent, each challenge's individual complexity has meant that, historically, they have been decoupled and studied individually. This paper creates a framework for batched pricing and matching in which pricing is seen as a meta-level optimisation over different possible matching decisions. Our key contributions are in developing a variant of the revenue-maximizing auction corresponding to the meta-level optimization problem, and then providing a scalable …

Transportation-Enabled Urban Services: A Brief Discussion, Hai Wang

Research Collection School Of Computing and Information Systems

Nearly 55% of the world's population lives in urban areas or cities, and is expected to rise above 70% over the coming decades. Rapid urbanization brings steadily more residents and a growing freelancing workforce into cities. The developments of city infrastructure and technologies—for instance, mobile location tracking and computing, autonomous and connected vehicles, wearable devices, robotics and robots, smart appliances, biometric authentication, various internet-of-things devices, and smart monitoring systems—are creating numerous opportunities and inspiring innovative and emerging urban services. Among these innovations, complex systems of urban transportation and logistics have embraced advances in technologies and, consequently, been significantly reshaped (Agatz …

Codem: Conditional Domain Embeddings For Scalable Human Activity Recognition, Abu Zaher Md Faridee, Avijoy Chakma, Zahid Hasan, Nirmalya Roy, Archan Misra

Research Collection School Of Computing and Information Systems

We explore the effect of auxiliary labels in improving the classification accuracy of wearable sensor-based human activity recognition (HAR) systems, which are primarily trained with the supervision of the activity labels (e.g. running, walking, jumping). Supplemental meta-data are often available during the data collection process such as body positions of the wearable sensors, subjects' demographic information (e.g. gender, age), and the type of wearable used (e.g. smartphone, smart-watch). This information, while not directly related to the activity classification task, can nonetheless provide auxiliary supervision and has the potential to significantly improve the HAR accuracy by providing extra guidance on how …

Learning To Solve Routing Problems Via Distributionally Robust Optimization, Jiang Yuan, Yaoxin Wu, Zhiguang Cao

Research Collection School Of Computing and Information Systems

Recent deep models for solving routing problems always assume a single distribution of nodes for training, which severely impairs their cross-distribution generalization ability. In this paper, we exploit group distributionally robust optimization (group DRO) to tackle this issue, where we jointly optimize the weights for different groups of distributions and the parameters for the deep model in an interleaved manner during training. We also design a module based on convolutional neural network, which allows the deep model to learn more informative latent pattern among the nodes. We evaluate the proposed approach on two types of wellknown deep models including GCN …

You Have Earned A Trophy: Characterize In-Game Achievements And Their Completions, Haewoon Kwak

Research Collection School Of Computing and Information Systems

Achievement systems have been actively adopted in gaming platforms to maintain players’ interests. Among them, trophies in PlayStation games are one of the most successful achievement systems. While the importance of trophy design has been casually discussed in many game developers’ forums, there has been no systematic study of the historical dataset of trophies yet. In this work, we construct a complete dataset of PlayStation games and their trophies and investigate them from both the developers’ and players’ perspectives.

Consensus Formation On Heterogeneous Networks, Edoardo Fadda, Junda He, Claudia J. Tessone, Paolo Barucca

Research Collection School Of Computing and Information Systems

Reaching consensus-a macroscopic state where the system constituents display the same microscopic state-is a necessity in multiple complex socio-technical and techno-economic systems: their correct functioning ultimately depends on it. In many distributed systems-of which blockchain-based applications are a paradigmatic example-the process of consensus formation is crucial not only for the emergence of a leading majority but for the very functioning of the system. We build a minimalistic network model of consensus formation on blockchain systems for quantifying how central nodes-with respect to their average distance to others-can leverage on their position to obtain competitive advantage in the consensus process. We …

Hu-Fu: Efficient And Secure Spatial Queries Over Data Federation, Yongxin Tong, Xuchen Pan, Yuxiang Zeng, Yexuan Shi, Chunbo Xue, Zimu Zhou, Xiaofei Zhang, Lei Chen, Yi Xu, Ke Xu, Weifeng Lv

Research Collection School Of Computing and Information Systems

Data isolation has become an obstacle to scale up query processing over big data, since sharing raw data among data owners is often prohibitive due to security concerns. A promising solution is to perform secure queries over a federation of multiple data owners leveraging secure multi-party computation (SMC) techniques, as evidenced by recent federation work over relational data. However, existing solutions are highly inefficient on spatial queries due to excessive secure distance operations for query processing and their usage of general-purpose SMC libraries for secure operation implementation. In this paper, we propose Hu-Fu, the first system for efficient and secure …

Faithful Extreme Rescaling Via Generative Prior Reciprocated Invertible Representations, Zhixuan Zhong, Liangyu Chai, Yang Zhou, Bailin Deng, Jia Pan, Shengfeng He

Research Collection School Of Computing and Information Systems

This paper presents a Generative prior ReciprocAted Invertible rescaling Network (GRAIN) for generating faithful high-resolution (HR) images from low-resolution (LR) invertible images with an extreme upscaling factor (64×). Previous researches have leveraged the prior knowledge of a pretrained GAN model to generate high-quality upscaling results. However, they fail to produce pixel-accurate results due to the highly ambiguous extreme mapping process. We remedy this problem by introducing a reciprocated invertible image rescaling process, in which high-resolution information can be delicately embedded into an invertible low-resolution image and generative prior for a faithful HR reconstruction. In particular, the invertible LR features not …

A Simple Data Mixing Prior For Improving Self-Supervised Learning, Sucheng Ren, Huiyu Wang, Zhengqi Gao, Shengfeng He, Alan Yuille, Yuyin Zhou, Cihang Xie

Research Collection School Of Computing and Information Systems

Data mixing (e.g., Mixup, Cutmix, ResizeMix) is an essential component for advancing recognition models. In this paper, we focus on studying its effectiveness in the self-supervised setting. By noticing the mixed images that share the same source images are intrinsically related to each other, we hereby propose SDMP, short for Simple Data Mixing Prior, to capture this straightforward yet essential prior, and position such mixed images as additional positive pairs to facilitate self-supervised representation learning. Our experiments verify that the proposed SDMP enables data mixing to help a set of self-supervised learning frameworks (e.g., MoCo) achieve better accuracy and out-of-distribution …

Metaformer Is Actually What You Need For Vision, Weihao Yu, Mi Luo, Pan Zhou, Chenyang Si, Yichen Zhou, Xinchao Wang, Jiashi Feng, Shuicheng Yan

Research Collection School Of Computing and Information Systems

Transformers have shown great potential in computer vision tasks. A common belief is their attention-based token mixer module contributes most to their competence. However, recent works show the attention-based module in transformers can be replaced by spatial MLPs and the resulted models still perform quite well. Based on this observation, we hypothesize that the general architecture of the transformers, instead of the specific token mixer module, is more essential to the model's performance. To verify this, we deliberately replace the attention module in transformers with an embarrassingly simple spatial pooling operator to conduct only basic token mixing. Surprisingly, we observe …

High-Resolution Face Swapping Via Latent Semantics Disentanglement, Yangyang Xu, Bailin Deng, Junle Wang, Yanqing Jing, Jia Pan, Shengfeng He

Research Collection School Of Computing and Information Systems

We present a novel high-resolution face swapping method using the inherent prior knowledge of a pre-trained GAN model. Although previous research can leverage generative priors to produce high-resolution results, their quality can suffer from the entangled semantics of the latent space. We explicitly disentangle the latent semantics by utilizing the progressive nature of the generator, deriving structure at-tributes from the shallow layers and appearance attributes from the deeper ones. Identity and pose information within the structure attributes are further separated by introducing a landmark-driven structure transfer latent direction. The disentangled latent code produces rich generative features that incorporate feature blending …

Shunted Self-Attention Via Multi-Scale Token Aggregation, Sucheng Ren, Daquan Zhou, Shengfeng He, Jiashi Feng, Xinchao Wang

Research Collection School Of Computing and Information Systems

Recent Vision Transformer (ViT) models have demonstrated encouraging results across various computer vision tasks, thanks to its competence in modeling long-range dependencies of image patches or tokens via self-attention. These models, however, usually designate the similar receptive fields of each token feature within each layer. Such a constraint inevitably limits the ability of each self-attention layer in capturing multi-scale features, thereby leading to performance degradation in handling images with multiple objects of different scales. To address this issue, we propose a novel and generic strategy, termed shunted selfattention (SSA), that allows ViTs to model the attentions at hybrid scales per …

Challenges For Inclusion In Software Engineering: The Case Of The Emerging Papua New Guinean Society, Raula Kula, Christoph Treude, Hideaki Hata, Sebastian Baltes, Igor Steinmacher, Marco Gerosa, Winifred Kula Amini

Research Collection School Of Computing and Information Systems

Software plays a central role in modern societies, with its high economic value and potential for advancing societal change. In this paper, we characterise challenges and opportunities for a country progressing towards entering the global software industry, focusing on Papua New Guinea (PNG). By hosting a Software Engineering workshop, we conducted a qualitative study by recording talks (n=3), employing a questionnaire (n=52), and administering an in-depth focus group session with local actors (n=5). Based on a thematic analysis, we identified challenges as barriers and opportunities for the PNG software engineering community. We also discuss the state of practices and how …

Understanding Crowdsourcing Requesters’ Wage Setting Behaviors, Kotaro Hara, Yudai Tanaka

Research Collection School Of Computing and Information Systems

Requesters on crowdsourcing platforms like Amazon Mechanical Turk (AMT) compensate workers inadequately. One potential reason for the underpayment is that the AMT’s requester interface provides limited information about estimated wages, preventing requesters from knowing if they are offering a fair piece-rate reward. To assess if presenting wage information affects requesters’ reward setting behaviors, we conducted a controlled study with 63 participants. We had three levels for a between-subjects factor in a mixed design study, where we provided participants with: no wage information, wage point estimate, and wage distribution. Each participant had three stages of adjusting the reward and controlling the …

Tatl: Task Agnostic Transfer Learning For Skin Attributes Detection, Duy M.H. Nguyen, Thu T. Nguyen, Huong Vu, Hong Quang Pham, Manh-Duy Nguyen, Binh T. Nguyen, Daniel Sonntag

Research Collection School Of Computing and Information Systems

Existing skin attributes detection methods usually initialize with a pre-trained Imagenet network and then fine-tune on a medical target task. However, we argue that such approaches are suboptimal because medical datasets are largely different from ImageNet and often contain limited training samples. In this work, we propose Task Agnostic Transfer Learning (TATL), a novel framework motivated by dermatologists’ behaviors in the skincare context. TATL learns an attribute-agnostic segmenter that detects lesion skin regions and then transfers this knowledge to a set of attribute-specific classifiers to detect each particular attribute. Since TATL’s attribute-agnostic segmenter only detects skin attribute regions, it enjoys …

Natural Attack For Pre-Trained Models Of Code, Zhou Yang, Jieke Shi, Junda He, David Lo

Research Collection School Of Computing and Information Systems

Pre-trained models of code have achieved success in many important software engineering tasks. However, these powerful models are vulnerable to adversarial attacks that slightly perturb model inputs to make a victim model produce wrong outputs. Current works mainly attack models of code with examples that preserve operational program semantics but ignore a fundamental requirement for adversarial example generation: perturbations should be natural to human judges, which we refer to as naturalness requirement. In this paper, we propose ALERT (Naturalness Aware Attack), a black-box attack that adversarially transforms inputs to make victim models produce wrong outputs. Different from prior works, this …

The Executive’S Guide To Getting Ai Wrong, Jerrold Soh

Asian Management Insights

This article explores how we see AI and argues that we mostly get it wrong. In the process, it explains the reasons backed by social science research on why we tend to get AI wrong and illustrates the dangers of doing so from a managerial and law-making perspective. Some readers may also find the article useful as a guide on how and when to manipulate portrayals of AI in your favour.

Engaging Undergraduate Students In An Introductory A.I. Course Through A Knowledge-Based Chatbot Workshop, Thomas Menkhoff, Ying Teo Lydia Qian

Research Collection Lee Kong Chian School Of Business

In this paper we share interim results of an ongoing mixed method evaluative study of 43 students enrolled in an elective course “Doing Business with A.I.” at the Lee Kong Chian School of Business (LKCSB), Singapore Management University. A key component of the course design is an experiential chatbot workshop that provides non-STEM students with an opportunity to acquire basic skills to build a chatbot prototype using the ‘Dialogflow’ program. The workshop and the experiential learning activity were designed to impart students with relevant knowledge and skills such as conversation and user-centric design know how and know why that are …

Neighbor-Anchoring Adversarial Graph Neural Networks (Extended Abstract), Zemin Liu, Yuan Fang, Yong Liu, Vincent W. Zheng

Research Collection School Of Computing and Information Systems

While graph neural networks (GNNs) exhibit strong discriminative power, they often fall short of learning the underlying node distribution for increased robustness. To deal with this, inspired by generative adversarial networks (GANs), we investigate the problem of adversarial learning on graph neural networks, and propose a novel framework named NAGNN (i.e., Neighbor-anchoring Adversarial Graph Neural Networks) for graph representation learning, which trains not only a discriminator but also a generator that compete with each other. In particular, we propose a novel neighbor-anchoring strategy, where the generator produces samples with explicit features and neighborhood structures anchored on a reference real node, …

Do Pre-Trained Models Benefit Knowledge Graph Completion? A Reliable Evaluation And A Reasonable Approach, Xin Lv, Yankai Lin, Yixin Cao, Lei Hou, Juanzi Li, Zhiyuan Liu, Peng Li, Jie Zhou

Research Collection School Of Computing and Information Systems

In recent years, pre-trained language models (PLMs) have been shown to capture factual knowledge from massive texts, which encourages the proposal of PLM-based knowledge graph completion (KGC) models. However, these models are still quite behind the SOTA KGC models in terms of performance. In this work, we find two main reasons for the weak performance: (1) Inaccurate evaluation setting. The evaluation setting under the closed-world assumption (CWA) may underestimate the PLM-based KGC models since they introduce more external knowledge; (2) Inappropriate utilization of PLMs. Most PLM-based KGC models simply splice the labels of entities and relations as inputs, leading to …

Chinese Idiom Understanding With Transformer-Based Pretrained Language Models, Minghuan Tan

Dissertations and Theses Collection (Open Access)

In this dissertation, I study the understanding of Chinese idioms using transformer-based pretrained language models. By ``understanding", I confine the topics to word embeddings learning, contextualized word representations learning, multiple-choice cloze-test reading comprehension and conditional text generation. Chinese idioms are fixed phrases that have special meanings usually derived from an ancient story. The meanings of these idioms are oftentimes not directly related to their component characters, which makes it hard to model them compared with standard phrases whose meanings are compositional. We initiate the work with studying idiom representations derived from pretrained language models, in particular, BERT. We adopt probing-based …

I'M Special But A.I. Doesn't Get It, Huei Huei Laurel Teo

Dissertations and Theses Collection (Open Access)

A growing body of management research on artificial intelligence (AI) has consistently shown that people innately distrust decisions made by AI and find such decision processes simply less fair compared to decisions made by humans. My dissertation adopts a different perspective to propose that aside from fairness concerns, AI decision methods trigger perceptions in people that their individual uniqueness has not be adequately considered and this has negative consequences for their psychological or subjective well-being.

By combining theories of uniqueness, individuality, power, and well-being, I develop five studies to provide empirical evidence that aversion to AI-mediated decisions also operates through …