Federated Graph Embedding for Efficient and Privacy Respecting COVID-19 Contact Tracing

Ethics of AI: Policy Guidance

Invited talk with the Asian Productivity Organization (APO), Tokyo, Japan.

CrowdFL: A Marketplace for Crowdsourced Federated Learning

Amid data privacy concerns, Federated Learning (FL) has emerged as a promising machine learning paradigm that enables privacy-preserving collaborative model training. However, there exists the need for a platform that matches data owners (supply) with model requesters (demand). We propose CrowdFL, a platform for facilitating the crowdsourcing of FL models. It provides support for client selection, model training, and reputation management, which are essential for the FL crowdsourcing operations. By implementing model training on actual mobile devices, we demonstrate that the platform improves model performance and training efficiency. To the best of our knowledge, it is the first platform to support crowdsourcing-based federated learning on edge devices.

Fairness in Design (FID)

As artificial intelligence (AI) becomes increasingly widely applied, societies have recognized the need for proper governance for its responsible usage. An important dimension of responsible AI is fairness. Due to the multi-faceted nature of the notions of fairness, it is challenging for AI solution designers to envision potential fairness issues at the design stage. In this project, we propose the Fairness in Design (FID) platform, an online collaborative design tool that aims to address the aforementioned issue. It provides AI solution designers with a workflow that allows them to surface fairness concerns, navigate complex ethical choices around fairness, and overcome blind spots and team biases.

Trust-based Open Collaborative Federated Learning

Today’s federated model training paradigm is based on an important assumption: After federated model training, all participants receive the same final model regardless of their contributions. This assumption may pose challenges to the adoption of federated learning, especially under the business-to-business settings. Suppose banks A, B and C want to collaboratively train a model to predict the creditworthiness of small and medium enterprises in a privacy-preserving manner. Bank A is significantly larger than B and C. If all of them receive the same final model under the current federated learning paradigm, Bank A which can contribute more to building a high quality model may hesitate for fear of benefiting other smaller banks and eroding its own market share. In this project, we address this challenge with the Fair and Privacy-preserving Deep Learning (FPPDL) approach. Unlike existing works which use money as incentives in federated learning, it uses the final model to incentivize collaborative behaviours. Specifically, federated learning participants will each receive a final model with performance reflecting their individual contribution to the federation.

crowded.sg: A Social Distancing Decision Support Platform for COVID-19

One of the most important problem for maintaining social distancing is to monitor the crowdedness of indoor and outdoor points of interest. In this paper, we report our experience developing and deploying an AI-empowered crowd counting platform to address this challenge. The platform relies on crowdsourced images and Unmanned Aerial Vehicles (UAVs) for data collection. Based on such data, state of the art deep learning algorithms in our AI engine performs analysis to produce the crowd counts. Initial deployment has been successfully carried out within the Nanyang Technological University (NTU) campus.

Fighting COVID-19 with AI: An Overview

A Multi-Agent Simulator of COVID-19 Propagation in Ride Sharing Systems

Order dispatch is an important area where artificial intelligence (AI) can benefit ride-sharing systems (e.g., Uber, Grab), which has become an integral part of our public transport network. In this project, we develop a multi-agent testbed to study the spread of infectious diseases through such a system. It allows users to vary the parameters of the disease and behaviours to study the interaction effect between technology, disease and people's behaviours in such a complex environment.

FedGame: A Multi-player Game for Studying Federated Learning Incentive Schemes

Federated Learning (FL) enables participants to "share" their sensitive local data in a privacy preserving manner and collaboratively build machine learning models. In order to sustain long-term participation by high quality data owners (especially if they are businesses), FL systems need to provide suitable incentives. To design an effective incentive scheme, it is important to understand how FL participants respond under such schemes. This video showcases FedGame - a multi-player game to study how FL participants make action selection decisions under different incentive schemes. It allows human players to role-play under various conditions. The decision-making processes can be analyzed and visualized to inform FL incentive mechanism design in the future.

Federated Learning powered Parkinson's Disease Risk Analytics based on Chinese Character Handwriting

Parkinson's disease (PD) is a debilitating neurodegenerative disorder. Early detection of the disease is important for effective treatment and can improve patients' quality of life. In order to achieve early detection, a population-level screening tool is required. In this video, we introduce a prototype that aims to capture early signs of PD symptoms from users' Chinese handwriting on mobile screens. It can be seamlessly infused into everyday use. It collects rich features of handwriting in the forms of images and time series data to support pattern analytics research that can assess users' PD risk. Personalized federated learning will be explored to form a privacy preserving machine learning backbone.

VSG: AI-empowered Visual Storyline Generator

Video editing is currently a highly skill- and time-intensive process. One of the most important tasks in video editing is to compose the visual storyline. in this project, we develop Visual Storyline Generator (VSG), an artificial intelligence (AI)-empowered system that automatically generates visual storyline based on a set of images and video footages provided by the user. It is designed to produce engaging and persuasive promotional videos with an easy-to-use interface. In addition, users can be involved in refining the AI generated visual storylines. The editing results can be used as training data to further improve the AI algorithms in VSG. The system has been deployed on the Aliyun platform.

FedCoin: A Blockchain-based P2P Payment System for Federated Learning Incentivization

As societies are increasingly aware of privacy issues related to artificial intelligence (AI) applications, federated learning (FL), which is a privacy-preserving distributed machine learning paradigm, emerged as a useful solution. To sustain FL ecosystems, it is important to provide fair incentives to FL clients. Shapley Value (SV) is a frequently used metric for distributing incentives fairly to FL clients. It is costly to compute. In this paper, we report FedCoin to address this problem. Instead of Proof-of-Work (PoW), participants of FedCoin create new blocks based on the proposed Proof of Shapley (PoSap) protocol. It can efficiently mobilize distributed computational resources through blockchain to help FL systems compute SVs. (http://demo.blockchain-neu.com/home).

Persuasive AI Companions for Active Independent Ageing

Most Societally Beneficial Video Award, IJCAI 2018.

PIDS

Electricity information tracking systems are increasingly being adopted across China. Such systems can collect real-time power consumption data from users, and provide opportunities for artificial intelligence (AI) to help power companies and authorities make optimal demand-side management decisions. In this project, we improve power utilization improvement in Shandong Province, China with a deployed AI application - the Power Intelligent Decision Support (PIDS) platform. Based on improved short-term power consumption gap prediction, PIDS uses an optimal power adjustment plan which enables fine-grained Demand Response (DR) and Orderly Power Utilization (OPU) recommendations to ensure stable operation while minimizing power disruptions and improving fair treatment of participating companies. This is a joint project with Dareway Software Co. Ltd and the Joint SDU-NTU Center for Artificial Intelligence Research (C-FAIR). Deployed in August 2018, the platform is helping over 400 companies optimize their power consumption through DR while dynamically managing the OPU process for around 10,000 companies. Compared to the previous system, power outage under PIDS through planned shutdown has been reduced from 16% to 0.56%, resulting in significant gains in economic activities. The project received the Innovative Application of Artificial Intelligence Award from AAAI in 2020.

AlgoCrowd

Years of rural-urban migration has resulted in a significant population in China seeking ad-hoc work in large urban centres. At the same time, many businesses face large fluctuations in demand for manpower and require more efficient ways to satisfy such demands. In this project, we develop AlgoCrowd, an artificial intelligence (AI)-empowered algorithmic crowdsourcing platform. Equipped with an efficient explainable task-worker matching optimization approach designed to focus on fair treatment of workers while maximizing collective utility, the platform provides explicable task recommendations to workers' personal work management mobile apps which are becoming popular, with the aim to address the above societal challenge. The platform has been deployed through a collaboration with WeBank and Better Life Commercial Chain Share Co. Ltd. The project received the Innovation Award from IJCAI 2019.

Spatial-Temporal Rebalancing (STR) Optimization for Shareable Bikes

Bike sharing systems are becoming a part of people's daily life in many cities. They provide commuters an environmentally friendly and low-cost transportation alternative for the first/last mile trips. However, these systems are facing a costly maintenance problem - rebalancing of shareable bikes among different docking stations. Ineffective rebalancing may result in undesirable operational situations including no-bike-to-pick (empty) and no-bike-to-drop (full). To address this challenge, we propose an efficient algorithm for the spatial-temporal rebalancing of shareable bikes within the budget based on queuing system dynamics. Compared to machine learning-based approaches, it provides a mechanism for system operators to incorporate their preferences into the rebalancing operation to optimize it without relying on the availability of large amounts of labelled usage data.

Algorithmic Crowdsourcing for Productive Aging

Today’s crowdsourcing combines the power of the Internet and the Crowd. However, it relies on the crowd to organize themselves and respond to task requests. Thus, it lacks efficiency and quality control.

At the Joint NTU-UBC Research Centre of Excellence in Active Living for the Elderly (LILY), our research focuses on the emerging field of inter-generational crowdsourcing. It aims to include the power of artificial intelligence (AI) into the picture to proactively motivate, organize, and manage a crowd from different age groups to collaborate so as to achieve more efficient utilization of their human resources.

Our patent pending algorithmic crowdsourcing technologies leverage on the trail of behaviour trajectory big data left behind by workers during their interactions to build knowledge about their trust relationships, their past performance, and their personal characteristics. Combined with the workflow templates provided by task requesters, we develop cutting edge technologies that efficiently utilize a limited budget, form workers into productive teams, and make quality-time-cost trade-offs when recommending tasks to workers.

The Silver Productivity App for Android Mobile Phones can be download from here.

ACDSS

The lack of systematic pain management training and support among primary care physicians (PCPs) limits their ability to provide quality care for pain patients. In this project, we built an Agent-based Clinical Decision Support System (ACDSS) to empower PCPs to leverage knowledge from pain specialists. The system analyses patients' health conditions related to pain, automatically generates treatment plans, and recommends a set of evidence-based pain cases to PCPs, with an embodied agent explaining the recommendations.

OIVIS

The Online Intelligent Visual Interactive System (OIVIS) project aims to produce an interactive live video broadcast tool empowered by advanced computer vision and technologies and human-computer interaction (HCI) designs. It can be applied to various live video broadcast and short video scenes to provide an interactive user experience. In the live video broadcast, the anchor can issue various commands by using pre-defined gestures, and can trigger real-time background replacement to create an immersive atmosphere. To support such dynamic interactivity, we implemented algorithms including real-time gesture recognition and real-time video portrait segmentation, developed a deep network inference framework, and a real-time rendering framework AI Gender at the front end to create a complete set of visual interaction solutions for use in resource constrained mobile.

The Agile Manager Game Platform

The Agile Manager is multi-agent game designed to help players experience the challenges of efficiently allocating tasks among a team of programmer with different capabilities. Behaviour data will be collected to form a dataset to help advance the research in human decision-making under uncertainty and resource constraints. You can learn more about the game and download the latest version from http://agilemanager.algorithmic-crowdsourcing.com/. A short video demonstrating the game play can be found below.

The Goal Net Designer Platform

The Goal Net Designer is an integrated multi-agent system design tool based on the Goal Net Methodology proposed by Dr. Zhiqi Shen from the School of Electrical and Electronics Engineering, NTU. Training videos for using the Goal Net Designer can be found here:

The SG50 Wish App

One of the key challenges of global population aging is the “empty nest” problem with many elderly people living alone at home, far away from their adult children. Interactive digital media (IDM) technologies can help to meet communication needs, bridge distances and make services much more accessible. There is great potential in using IDM technologies to improve the quality of life of the elderly.

Residents can use the app to make SG50 wishes through photos or text. The app is equipped with Augmented-Reality technology which allows it to detect SG50 logos to promote inter-generation interactions and multiculturalism. SG50 wishes taken by the app can then be easily shared with neighbours and friends. The app can also be used as a platform for social interaction revolving around people’s daily lives, and enables people to communicate feelings that they are either too busy or too shy to convey in person.

The platform has been launched by Minister for Communications and Information, Dr Yaacob Ibrahim (https://simisaialsosg50.wordpress.com/2015/02/17/new-app-launched-to-collect-singaporeans-wishes-for-sg50/).

The project set a new Singapore Record for the Most Number of Wishes Received from an Augmented Reality App, achieving over 50,000 wishes by 8th Aug, 2015.

The app can be downloaded from here.

The platform website can be accessed from here.