The day featured three invited talks on various topics. The first morning talk was given by Zhenhua Dong, who gave us comprehensive perspectives on biases in recommender system, covering both out of distribution and unfairness. The second morning talk was given by Wenxuan Zhang, who took us through the several interesting works in advancing multilingual large language models. The afternoon talk was given by Shonali Krishnaswamy, who gave a captivating sharing of her successful AI startup journey.

For the first time, we ran the presentations along two tracks: Research Track, managed by Yuan Fang and Cheng Long, as well as Applied Data Science Track, managed by Yong Liu and Wee Siong Ng. The research track featured seven oral presentations and eleven posters, while the applied data science track featured two oral presentations.

At the end of the day, we announced several awards based on audience feedback, but the reality was that all attendees won, because we all gained insights and friendships.

Date: September 22, 2023

Time: 4pm to 5pm

Venue: SMU School of Computing and Information Systems 2, Seminar Room B1.2

Abstract

In this talk, we delve into the fascinating world of large-scale recommender systems, exploring three key dimensions: Conformity, Exploration and Context Awareness. We present innovative approaches to break free from conformity bias, infuse exploration to capture evolving user interests and incorporate context awareness for more personalized recommendations. Through rigorous experimentation and real-world deployments on platforms like Facebook Watch, we unveil how these methodologies revolutionize recommendation systems. Join us to uncover the secrets behind unlocking personalization in large-scale recommendations, offering profound insights that bridge the gap between user engagement and system performance.

Speakers

• Khushhall Chandra Mahajan is a Senior Machine Learning Engineer and Researcher working in recommendation systems. Presently, he is an integral part of the Video Recommendations team at Meta, where he plays a pivotal role in crafting cutting-edge ML algorithms. These algorithms directly enhance the accuracy and impact of video recommendations used by billions of users worldwide. His work powers the recommendation engine for Facebook Watch and Reels, reaching over 1.25 billion users monthly. Prior to this role, he was in the Ads team, developing innovative Ads solutions that drove top-line revenue for the company. He has published several research papers in the domain of machine learning and recommendation systems. His research interest focuses on exploration & diversity in recommendation system and realtime ranking. Additionally he has served on the program committee of various top-tier international machine learning conferences such as NeurIPS, ICLR, ECIR, AISTATS, etc and organizer of the VideoRecsys workshop in ACM RecSys conference. As a fascinating side note, his innovation extends beyond machine learning; he has also developed the Swarachakra Bangla Android keyboard, which has garnered over a million downloads.

• Amey Porobo Dharwadker works as a Machine Learning Engineering Manager at Meta, where he leads the Facebook Video Recommendations Core Ranking team responsible for developing personalization models used by billions of users around the world. His work has been instrumental in driving the remarkable growth in active users for Facebook, by powering the success of Facebook Watch and Reels, reaching more than 1.25 billion monthly users. Prior to that, he delivered significant user engagement and revenue growth for Facebook through advancements in News Feed and Ads Machine Learning. He has published several research papers in the fields of large-scale recommendation systems and machine learning. He also actively participates as a program committee member for top-tier AI venues including AISTATS, AAAI, IJCAI, ECIR, CIKM, etc. and organizes the VideoRecsys workshop at ACM RecSys conference. He also serves on the juries of renowned global technology competitions, including the Edison Awards and Globee Information Technology Awards.

Photos

Thank you Amey and Khushhall for taking the time to share your work on recommender systems with the SIGKDD community in Singapore.

Below is a brief photo-story of the event.

We’re certainly excited to do this all over again. Keep a look out for our next event!

In this seminar by KDD.SG held on April 26, 2023, we featured three experts on graphs.

Talk #1: Title: Efficient Graph Representations for Dynamic Processing on GPUs: Leveraging Dynamic Data Structures for Adaptability and Versatility

Abstract: Dynamic graph processing has become increasingly important in various fields, such as social network analysis, disease surveillance, and transaction monitoring, where graphs evolve over time or need to be analyzed in real-time. Graphics Processing Units (GPUs) offer massive parallelism and high computational power, making them an ideal choice for processing large-scale dynamic graphs. However, efficient processing of these dynamic graphs on GPUs demands adaptive and versatile graph representations. In this talk, we delve into the design of innovative graph representations that enable efficient dynamic graph processing on GPUs. We will discuss the characteristics of our graph representation, which utilizes dynamic data structures to cater to different update requirements. By highlighting the importance of adaptability and versatility in graph representations, attendees will gain valuable insights into optimizing dynamic graph processing on GPUs and be better equipped to enhance their own graph-based applications.

Bio: Yuchen Li is an assistant professor with the School of Computing and Information Systems, Singapore Management University (SMU). He received double BSc degrees in applied math and computer science (both with first-class honors) and a Ph.D. degree in computer science from the National University of Singapore (NUS), in 2013 and 2016, respectively.

Talk #2: Dense subgraph mining: problems, algorithms and applications

Abstract: A graph represents a set entities as vertices and the relations among the entities as edges and is used for modelling data in many applications. A subgraph of vertices in a graph, which are connected via many edges, is called a dense subgraph and often conveys meaningful knowledge of the graph. Finding dense subgraphs in a graph has a wide range of applications including correlation mining, fraud detection, e-commerce, bioinformatics, frequent pattern mining, and community detection, etc. In this talk, I will first introduce some representative applications, density measures and problems of dense subgraph mining. I will then present several of our recent studies of finding dense subgraphs in bipartite graphs and uncertain graphs and their applications in fraud detection, community search, and fake news detection. Finally, I will discuss some potential future work directions.

Bio: LONG Cheng is currently an Assistant Professor at the School of Computer Science and Engineering (SCSE), Nanyang Technological University (NTU). He got the PhD degree from Hong Kong University of Science and Technology (HKUST) in 2015. His research interests are broadly in data management and data mining.

Talk #3 Low-Resource Learning on Graphs
Abstract: Graph structures are ubiquitous in various domains, ranging from social networks and e-commerce platforms to transportation and biological systems. On these graphs, various graph-based analytics and mining tasks exist, many of which can be cast as instances of link prediction, node classification, and graph classification. Moving away from manual feature engineering, graph neural networks (GNN) have witnessed widespread success in various application scenarios due to their ability to learn powerful graph representations automatically. However, their success is often dependent on the availability and quality of graph structures and labeled data, without which their performance can suffer. In this talk, we explore alternative learning paradigms different from the traditional supervised learning paradigm, specifically addressing two types of low-resource scenarios on graphs: structure scarcity and label scarcity. We will first provide an overview of low-resource problems and methods on graphs, and then introduce some of our representative works on these problems.

Bio: Dr. Yuan Fang is an Assistant Professor at the School of Computing and Information Systems at Singapore Management University (SMU). He was previously a data scientist at DBS Bank and a research scientist at A*STAR. His research interests revolve around graph-based learning and its applications in recommender systems, social network analysis, and bioinformatics.

On January 26, 2022, SIGKDD Singapore held an online lunch-time seminar featuring two talks on KDD for a Safer Online Social Space by two experts in the field.

Talk#1: Perils and Promises of Automated Hate Speech Detection

Abstract: Online hate speech is an important issue that breaks the cohesiveness of online social communities and even raises public safety concerns in our societies. Motivated by this rising issue, researchers have developed many traditional machine learning and deep learning methods to detect hate speech in online social platforms automatically. This talk aims to introduce the pressing problem of online hate speeches and discuss the automated hate speech detection methods. Specifically, we will discuss the various NLP approaches for hate speech detection and highlight the potential future research directions such as multimodal and multilingual hate speech detection.

Bios: Roy is an Assistant Professor at the Information Systems Technology and Design Pillar, Singapore University of Technology and Design. He is a faculty of the transformative Design and Artificial Intelligence program. His research lies in the intersection of data mining, machine learning, social computing, and natural language processing. He is leading the Social AI Studio, a research group that focuses on designing the next-generation social artificial intelligence systems. He has published in top-tier venues in data mining and computation linguistics domains. He serves in the program committees of multiple top conferences. He is currently part of the editorial board for the Social Network Analysis and Mining journal.

Talk#2: Rumor Detection with Generative Adversarial Learning

Abstract: Online rumors can cause devastating outcomes to individuals and society. Analysis shows that the widespread of rumors typically results from deliberate promotion of uncredited information aiming to shape the public opinions. On the other hand, fact-checking currently follows investigative journalism requiring significant amount of time and manual effort, which cannot keep the pace of generation of various rumors on a daily basis. In this talk, I will introduce automatic approaches, techniques and new development for combating online rumors in social media from the perspective of natural language processing and adversarial learning. I will also discuss the characterization of online rumors including their linguistic, temporal and propagational features and dynamics, together with some takeaways from past and ongoing research.

Bio: Wei Gao is currently an Assistant Professor of Computer Science at Singapore Management University. His research generally interests natural language processing, information retrieval, artificial intelligence and social computing. Currently he has been working on the topic of rumor detection and computational fact checking. His publications appear in the major international venues including ACL, AAAI/IJCAI, SIGIR, WSDM, WWW, ACM TOIS, IEEE TKDE, etc. He broadly serves the top conferences and leading journals in his relevant field. He is an Associate Editor of ACM TALLIP and a member of standing review committees of Transactions of the ACL and Computational Linguistics Journal.

Dr Markus Kirchberg represented OpenMined to open the session 

The first talk on data privacy in machine learning was given by Assistant Professor Reza Shokri, who is NUS Presidential Young Professor.

Reza introduced his research on data privacy and trustworthy machine learning

The second talk on tests and metrics to evaluate machine learning model explanations was given by Naresh Rajendra Shah, who is Co-founder and CTO at Untangle AI.

Naresh on how we could evaluate explanation for a machine learning model

We thank both speakers as well as OpenMined for what turned out to be a mind-opening experience for all the attendees.

May 29, 2019 | 6pm to 8.30pm | WeWork 71 Robinson

Abstract

Cloud Computing has become a mainstay of IT industry in recent years. Many variations and flavors had since developed, and they can be quite confusing to those who are unfamiliar. Many chose to follow suit simply because it is a growing trend, but they may not be making the most appropriate decisions for their businesses if the intricacies aren’t studied thoroughly.

Cloud Computing in general can be viewed as a 3 layer architecture where, IaaS (infrastructure as a service) sits at the bottom, PaaS (platform aaS) in the middle, and SaaS (software aaS) on top.

Being the foundation layer, IaaS is critical in dictating an efficient and effective infrastructure that powers today’s businesses. In order to provision, deploy, and manage IaaS appropriately, specific terminology was developed to illustrate some of the most common deployment models, including private cloud, public cloud, hybrid cloud, and multi-cloud.

While public cloud, private cloud, and even hybrid cloud have been popularized, business needs continue to evolve and technology continue to advance. Today, Hybrid multi-cloud is becoming a growing trend in cloud computing. Deploying a cloud infrastructure that combines private on-premise data centers and multiple public cloud providers (hybrid multi-cloud) is daunting, but businesses continue to flock to this model. In his talk, Michael will discuss current industry trends and consumptions surrounding hybrid multi-cloud, implications of business and technology impact, and illustrate through a use case the need to embrace what is to come.

Speaker

Michael Tsang currently serves as Senior Director of Channel Management APAC at Equinix, the leading data center provider in the world.

Prior to joining Equinix, Michael led engineering and research efforts at Hulu for 1 year, in areas such as recommendation, prediction, and pattern recognition.

Before Hulu, Michael spent 3.5 years with Alibaba, responsible for driving market entry and business growth in North and South America regions. He also led and completed Alibaba Cloud’s European data center strategy and owned Hong Kong’s partner management.

Prior to that Michael was with Microsoft for 16 years in various engineering roles including development and testing; gaining experience in operating systems, business rule engines, video rendering and transcoding, online advertising, and big data stream computing. Prior to Microsoft, he served at Hewlett-Packard in LaserJet and network management products.

Date/Time Wednesday 29 May 2019, beginning with a light reception at 6pm

Venue WeWork 71 Robinson Road Singapore 068895

RSVP on our Meetup.

See you there!

Agenda of the Seminar

Prof Jie Tang from Tsinghua University summarized several works on using deep learning for deriving graph representations, distilled their fundamental concepts, and described several generalizations and extensions on such models.

Jie describing the process of learning embeddings of nodes from a graph, in this case via employing Skip-gram technique on random walk paths on the graph

Jie discussing how NSGCN, a graph convolutional network with partial nodes’ features, would be much more efficient than ordinary GCN, yet obtaining close performance

In turn, Prof Michalis Vazirgiannis from Ecole Polytechnique described his recent work on using deep learning to derive representations for sets.

Michalis introducing a recent work on using neural networks to learn set representations

Michalis providing an overview of several related works on deep learning for graphs emerging from his lab

The audience was engaged, and the ensuing questions-and-answers were lively with a number of relevant questions. Overall, it was a very informative and inspiring session that certainly amplified our understanding of deep learning and neural networks.

Talks

• “Graph Neural Networks and Applications” by Jie Tang of Tsinghua University
• “Rep the Set: Neural Networks for Learning Set Representations” by Michalis Vazirgiannis of Ecole Polytechnique

Date: April 25th 2019
Time: 4.00pm to 6.30pm, beginning with light refreshments
Venue: SIS Seminar Rm B1-1, SMU School of Information Systems, 80 Stamford Road, Singapore 178902

RSVP on our Meetup

Talk#1: Graph Neural Networks and Applications

Graph Neural networks (GNNs) and their variants have generalized deep learning methods into non-Euclidean graph data, bringing a substantial improvement on many graph mining tasks. In this talk, I will revisit graph convolutional networks and investigate how to improve their representation capacity. We discover that the performance of GNNs can be significantly improved with several simple and elegant refinements on the neighborhood aggregation and network sampling steps. Importantly, we show that some of the most expressive GNNs, e.g., the graph attention network, can be reformulated as a particular instance of our models. Extensive experiments on different types of graph benchmarks show that our proposed framework can significantly and consistently improve the graph classification accuracy when compared to state-of-the-art baselines.

Speaker

Jie Tang is a Full Professor and the Vice Chair of the Department of Computer Science and Technology at Tsinghua University. His interests include data mining, social networks, knowledge graph, machine learning, and artificial intelligence. He has been visiting scholar at Cornell University, Hong Kong University of Science and Technology, and Southampton University. He has published more than 300 journal/conference papers and holds 20 patents. His papers have been cited by more than 12,000 times. He served as PC Co-Chair of CIKM’16, WSDM’15, Associate General Chair of KDD’18, and Acting Editor-in-Chief of ACM TKDD, Editors of IEEE TKDE/TBD and ACM TIST. He leads the project AMiner.org for academic social network analysis and mining, which has attracted more than 10 million independent IP accesses from 220 countries/regions in the world. He was honored with the UK Royal Society-Newton Advanced Fellowship Award, CCF Young Scientist Award, NSFC for Distinguished Young Scholar, and KDD’18 Service Award.

Talk#2: Rep the Set: Neural Networks for Learning Set Representations

In several domains, data objects can be decomposed into sets of simpler objects. It is then natural to represent each object as the set of its components or parts. Many conventional machine learning algorithms are unable to process this kind of representations, since sets may vary in cardinality and elements lack a meaningful ordering. In this paper, we present a new neural network architecture, called RepSet, that can handle examples that are represented as sets of vectors. The proposed model computes the correspondences between an input set and some hidden sets by solving a series of network flow problems. This representation is then fed to a standard neural network architecture to produce the output. The architecture allows end-to-end gradient-based learning. We demonstrate RepSet on classification tasks, including text categorization, and graph classification, and we show that the proposed neural network achieves performance better or comparable to state-of-the-art algorithms

Speaker

Dr.  Vazirgiannis is a Professor at LIX, Ecole Polytechnique in France. He has conducted research in Frauenhofer and Max Planck-MPI (Germany), in INRIA/FUTURS (Paris). He has been a teaching in AUEB (Greece), Ecole Polytechnique, Telecom-Paristech, ENS (France), Tsinghua, Jiaotong Shanghai (China) and in Deusto University (Spain). His current research interests are on deep and machine learning for Graph analysis (including community detection, graph classification, clustering and embeddings, influence maximization), Text mining including Graph of Words, deep learning for word embeddings with applications to web advertising and marketing, event detection and summarization. He has active cooperation with industrial partners in the area of data analytics and machine learning for large scale data repositories in different application domains. He has supervised twenty completed PhD theses. He has published three books and more than a 200 papers in international refereed journals and conferences and received best paper awards in ACM CIKM2013 and IJCAI2018. He has organized large scale conferences in the area of Data Mining and Machine Learning (such as ECML/PKDD) while he participates in the senior PC of AI and ML conferences – such as AAAI and IJCAI, He has received the ERCIM and the Marie Curie EU fellowships, the Rhino-Bird International Academic Expert Award by Tencent and since 2015 he leads the AXA Data Science chair.