External Research Grants

Under the Fujitsu-SMU Urban Computing and Engineering (UNiCEN) Corp Lab, SMU has undertaken the Digital Platform Experimentation (DigiPlex) project with Fujitsu. The project was carried out using the Digital Annealer (DA), a quantum inspired-technology inspired by Fujitsu. Through the DigiPlex project, certain challenges in solving constrained optimization problems using such technology, and promising methods on tuning of the underlying model parameters to improve run time performance, have been identified. This project aims at developing hyper parameter tuning methodology, machine learning techniques, operations research algorithms, and software tools to enhance quantum-inspired techniques for solving large scale real-world combinatorial optimization problems.

This project aims to develop Fast-Adapted Neural Networks (FANNs) and provide specific solutions to equipping the Advanced Manufacturing and Engineering (AME) systems with FANNs. It considers a wide range of AME application examples such as visual inspection of new product parts and automated identification of product defects. It will improve the yield rate and reduce manufacturing costs, when FANNs-based devices are widely installed in the design, layout, fabrication, assembly, and testing processes of production lines. This research is supported by A*STAR under its AME YIRG Grant (Project No. A20E6c0101).

This collaboration with Coleridge Initiative aims to develop entity linking models that automatically identify datasets used in research publications to perform context-rich search and recommendation on research repositories.

Smart systems are increasingly dependent on machine learning frameworks for their feature implementation. These frameworks are built on top of many third-party libraries, which depend on many others. Simply trusting and reusing a framework poses a security risk as the framework and third-party libraries it depends on can contain exploitable vulnerabilities. To mitigate this risk, this project will create an advanced solution that identifies vulnerabilities in popular machine learning frameworks.

This research collaboration with Zhejiang University seeks to unlock the power of large software data stored in open software repositories for automating three common software development tasks: coding (code completion), commenting (code summarization), and identification of software defects (defect prediction).

Singlish is an English Creole language used in Singapore and it is evolving. New words have been introduced from time to time making it difficult to track the development of this language. With Singlish being used largely in informal conversations, it is extremely challenging for any linguistic expert to keep track of its changes. It is therefore necessary to combine machine learning and human expertise in a crowdsourcing approach to construct and maintain a Singlish dictionary at speed and scale, with reasonable quality. This project will use a combination of machine learning and human efforts in a crowdsourcing approach to learn Singlish words so as to build some sort of Singlish dictionary.

Software developers often make use of a method called satisfiability modulo theories (SMT) for program testing, analysis and verification. Unfortunately, SMT has its limitations as it is ineffective for large software that are usually more complicated. As such, the collaborative project with Huawei, aims to develop new constraint solving techniques based on optimization to improve the effectiveness and efficiency of testing and formal verification of software, so that it can be used on large software as well.

The K-EMERGE research programme, funded by the RIE2020 Advanced Manufacturing and Engineering Programmatic Grant, proposes a knowledge-based AI approach, complemented by advances in deep-learning NLP methods, to address the need for AI systems that are able to perform deep inference for expert-level diagnosis, explanation, instruction, and decision aiding in the context of complex physical systems. As part of the K-EMERGE programme, this project undertaken by SMU aims to develop computational models and technologies for representation, modelling and learning of domain knowledge extracted from text-based technical documents.

Through usage of city-scale commuting data (e.g., public transport records captured by smart card EZLink) to perform multimodal data analysis, this project seeks to answer following questions:

• Question 1: when and where do people commute (related to trip prediction)
• Question 2: how they commute (related to recovery of the exact routes taken by commuters inside the MRT network)
• Question 3: why they commute (related to inference of trip purposes)

A novel parallel framework named CONQUEROR is proposed in this project to support large scale concurrent graph query processing. The framework is built upon the popular heterogeneous architecture, which consists of both CPUs (central processing units) and GPUs (graphics processing units), and they aim to develop a set of novel parallel approaches to fully harness the unique characteristics of the heterogenous platform for processing millions of graph queries concurrently.