DR-NTU (Digital Repository of NTU)
Not a member yet
116018 research outputs found
Sort by
Graph connectivity with fixed endpoints in the random-connection model
No full textWe consider the count of subgraphs with an arbitrary configuration of
endpoints in the random-connection model based on a Poisson point process on
Rd. We present combinatorial expressions for the computation of the
cumulants and moments of all orders of such subgraph counts, which allow us to
estimate the growth of cumulants as the intensity of the underlying Poisson
point process goes to infinity. As a consequence, we obtain a central limit
theorem with explicit convergence rates under the Kolmogorov distance, and
connectivity bounds. Numerical examples are presented using a computer code in
SageMath for the closed-form computation of cumulants of any order, for any
type of connected subgraph and for any configuration of endpoints in any
dimension d>=1. In particular, graph connectivity estimates, Gram-Charlier
expansions for density estimation, and correlation estimates for joint subgraph
counting are obtained.Ministry of Education (MOE)Published versionThis research is supported by the Ministry of Education, Singapore, under its Tier 2 Grant MOE-T2EP20120-0005
Bayesian neural networks with possibility theory
No full textIn this paper, we introduce the use of possibility theory into bayesian neural networks to attempt to quantify uncertainty of neural network predictions. While there exist other methods which have their own quantification, they face certain constrains such as scalability or the use of parametric assumption. In the method we propose, we aim to rectify these constrains while still maintaining reasonable results for uncertainty quantification.Bachelor's degre
Resilience modelling and assessment of maritime transportation systems
No full textMaritime transportation systems (MTSs) play a crucial role in global trade and economic development by facilitating the movement of goods and commodities across vast distances, thereby connecting producers and consumers worldwide. However, MTSs face disruptions from various sources. To enhance the resilience of MTSs—their ability to cope with these disruptions—this thesis aims to conduct modelling and assessment studies on their resilience.
In terms of research methodology, considering data availability, both data-driven methods and expert opinion extraction are employed. For data that is difficult to obtain, this thesis proposes a novel method to extract expert opinions based on Dempster-Shafer evidence theory and hesitant fuzzy linguistic terms. To establish causal relationships between variables, Bayesian network methods are also investigated. For available data, detailed data cleaning, analysis, and simulation model construction are conducted for different components. Specifically, for shipping networks, a method for detecting port communities based on the Infomap algorithm is proposed.
In the application of this research, the three most critical components of MTSs—waterways, ports, and shipping networks—were selected for analysis. The main contributions of these applications and analyses can be summarized as follows.
Waterway Transportation Channel (WTC) Resilience Study: A discrete-event simulation model is proposed to quantify WTC resilience by collecting Automatic Identification System data from the Yangtze Estuary Deepwater Channel. The model simulates ship operations within this specific waterway, an area currently limited in the literature. Different scenarios, such as ship delay and ship load, are designed to test the system’s performance. Based on the analysis of experimental results, several practical suggestions are provided for stakeholders to aid in WTC risk management. By considering ship load, ship delay, and recovery cost, a composite resilience indicator is proposed to help managers design rescue plans. Using real accident data, different rescue scenarios are proposed and compared, with simulation results demonstrating the effectiveness of the new indicator.
Port Resilience Study: A circular four-stage method is proposed to study port resilience, incorporating a port resilience assessment model using Bayesian networks. This model categorizes resilience strategies into six metrics (robustness, redundancy, visibility, flexibility, agility, recovery) to assess resilience capabilities (readiness and response), aiding port managers in distinguishing the specific functions and effectiveness of different strategies. Major disturbances affecting ports are summarized, with the Shanghai Yangshan Deepwater Port in China used as a case study for quantifying port resilience. Sensitivity analysis validates the model and compares the effects of resilience strategies and metrics on resilience capacities. Forward and backward inferences identify pathways to achieving a resilient port system. Key conclusions from the case study include: natural disasters are major disruptors; automated terminals exhibit higher overall resilience; and strategies enhancing visibility and recovery are crucial for readiness and response, respectively.
Shipping Network Resilience Study: An enhanced disruption simulation model is proposed and a novel research perspective: port communities is introduced. The simulation model integrates cascading failure and recovery mechanisms, incorporates ship behaviour during disruptions, and introduces a temporal dimension to track the network's evolution. Port community-to- community connections provide a clearer and more holistic perspective. Using Infomap algorithm, port communities are identified based on transportation direction, capacity, and geographic proximity, resulting in a decentralized and balanced structure while preserving GCSN's scale-free and small-world properties. Simulations of various disruption scenarios and recovery strategies yielded optimized key parameters and practical recommendations. For instance, the optimal distance threshold for detecting port communities is 300 km. Additionally, weak correlations between alternative port numbers and community size/throughput (0.17, 0.246) underscore the need for geographically balanced distribution and reduced reliance on single ports.
Overall, this thesis presents resilience modelling and assessment studies of key components within MTSs. The findings offer practical recommendations for strengthening the resilience of maritime operations, ensuring the system's capability to withstand and recover from disruptions. Moreover, the research methodology provided in this study demonstrates versatility and applicability across various contexts, offering a valuable framework for future investigations into the resilience of complex transportation networks.Doctor of Philosoph
Highly controllable motion generation model
No full textText-To-Motion generation has emerged as a promising area of research in deep learning,
with potential applications in video games, animation and virtual reality systems.
However, the adoption of these technologies is still limited due to the predefined
skeletal prior. Thus, manual effort is required to rig the desired target meshes with a
compatible skeleton. On the other hand, recent advancements in 3D-Human generation
have demonstrated the capability to produce detailed and realistic 3D character models
from textual inputs. The gap between motion and 3D-human generation is a compelling
area of research. A pipeline that can automate the transfer of motion to generated 3D-
human model will significantly simplify the workflow of generating 3D animations for
the laypersons. This project reviews the state-of-the-art (SOTA) approaches in motion
and 3D-Human generation, as well as methods in ensuring seamless compatibility
between them. We propose a pipeline that integrates both models to enable automated
and user-friendly workflows for creating 3D animations whilst ensuring compatibility
with popular 3D software platforms like Unreal Engine and Blender.Bachelor's degre
The poetics of contemporary Muslim immigrant fiction
No full textWhen looking at categories for the study of “World Literature”, the geographical situatedness of a novel’s particular setting or an author’s (birth)place of origin is often taken as a reference. Yet, mobility studies have shown the value of looking beyond the frames of Arab-American or South Asian literature. In this thesis, I posit the value of considering “Muslim immigrant fiction” as a category to better understand, explore, and compare the thematic and aesthetic influences of Muslim representations in world literature—as opposed to merely in isolated regions such as South Asia or amongst distinct ethnic groups like Arabs. Muslim fiction is a burgeoning field, both in academia and amongst contemporary readers; yet, most novels that fit within this genre tend to be examined as immigrant narratives or under the broader category of world literature. Despite growing interest in the field, Muslim fiction remains mostly amorphous and inadequately defined, with gaps in the scholarship. I proffer that this ambiguity of defining Muslim fiction takes its cue from the current anxieties that shape and complicate a Muslim’s definition of their own identity.
As such, I strive to make a meaningful contribution to scholarly conversations by first proposing a working definition of the genre of “Muslim immigrant fiction”, drawing on existing scholarship and as shaped by the subsequent analyses of my selected case studies, Mohsin Hamid’s Exit West, Zoulfa Katouh’s As Long As The Lemon Trees Grow, and Marjan Kamali’s The Stationery Shop. I build on the work of scholars such as Frederick Luis Aldama and Sue-Im Lee to move beyond the postcolonial tendency of ethnographic transparency by putting forth a framework to analyse the poetics of Muslim immigrant fiction.
This paper adopts Jonathan Culler’s definition of “poetics” as “the characteristic techniques, compositional habits, and ways of treating subjects in the literary practice under consideration.” My paper thus examines the poetics of Muslim immigrant fiction by exploring the various characteristic techniques and compositional habits—in the form of narrative features and thematic ideas—used to represent the Muslim experience in these narratives. In doing so, I hope to use this framework to:
a) identify a range of characteristic narrative features and themes for analysing Muslim subjectivity in the genre of Muslim immigrant fiction; and
b) establish the productivity of reading a group of texts through this framework and to articulate the interpretive payoffs of doing so.
I posit the value of tracing the poetics of Muslim immigrant fiction by focusing on the aesthetics of estrangement, migrational trauma, and discourses of resistance.Master's degre
Designable excitonic effects in van der Waals artificial crystals with exponentially growing thickness
No full textWhen disassembled into monolayers from their bulk crystals, two-dimensional (2D) transition metal dichalcogenides (TMDCs) exhibit exotic optical properties dominated by strong excitonic effects. Reassembling 2D TMDC layers to build bulk excitonic crystals can significantly boost their optical performance and introduce emerging functionalities toward optoelectronic and valleytronic applications. However, maintaining or manipulating 2D excitonic properties in bulk structures or superlattices is challenging. Herein, we developed a method to precisely construct m∙2N-layer artificial excitonic crystals with only a number N of stacking operations (m denotes the layer number of the initial material unit), referred to as the "2^N method". We successfully fabricated a millimeter-scale weakly coupled 16-layer MoS2 single crystal with zero interlayer twist angle, which retains monolayer-like exciton properties and exhibits remarkable enhancements up to 643% and 646% in their absorption and photoluminescence (PL) features, respectively. Moreover, we created a WSe2/(MoS2/WSe2)3/MoS2 superlattice starting from monolayer WSe2 and MoS2, which demonstrated an intensity increase of up to 400% in quadrupolar interlayer exciton (IX) emission as compared to dipolar IXs in its bilayer counterpart. Our work shows a promising approach for the design and bottom-up fabrication of excitonic crystals, promoting the exploration of excitonic physics in complex van der Waals (vdW) structures and their applications in optoelectronic devices.Ministry of Education (MOE)Nanyang Technological UniversityPublished versionW.X. gratefully acknowledges support from the National Key R & D Program of China (2022YFA1204701), the Fundamental Research Funds for the Central Universities in China (2024300393), the National Natural Science Foundation of China (22333004, 22173044), the Natural Science Foundation of Jiangsu Province (BK20220121). Z.H. gratefully acknowledges support from the National Key R & D Program of China (2022YFA1204301), the National Natural Science Foundation of China (62405132), the Natural Science Foundation of Jiangsu Province (BK20241227). R.S. gratefully acknowledges funding support from the Singapore Ministry of Education via AcRF Tier 2 grant (MOE-T2EP50222-0008), AcRF Tier 3 grant (MOE-MOET32023-0003) “Quantum Geometric Advantage”, Tier 1 grant (RG80/23) and Nanyang Technological University via a Nanyang Assistant Professorship start-up grant
Multi-agent soft actor-critic aided active disturbance rejection control of dc solid-state transformer
No full textThe dc solid-state transformer (dcSST) plays a vital role in interconnecting diverse dc sources and loads in dc microgrids. However, output voltage regulation and submodule power balance control have always been two essential control challenges of the dcSST. To address these challenges, this article proposes a multiagent soft actor-critic-based active disturbance rejection control (MASAC-ADRC) method. The ADRC method is used for uncertainty estimation and compensation in modular dcSST. Specifically, the controller gains of the ADRC method are optimized adaptively by the MASAC method, thus enhancing the adaptability to changing conditions from the environment. Compared with the existing controller gain optimization methods, the proposed method does not rely on predetermined datasets. Instead, it provides a tailored strategy, employing a neural network to map optimal ADRC parameters from the measured states. Through the incorporation of diverse environmental scenarios, encompassing variant input voltage and output power, the MASAC-ADRC method achieves superior dynamic performance. Experimental validation underscores the efficacy of the proposed algorithm, showcasing its superiority over alternative approaches. The proposed method yields enhancements exceeding 50% in dynamic performance metrics such as overshoot, settling time, and mean square error when compared with existing methods.National Research Foundation (NRF)Nanyang Technological UniversitySubmitted/Accepted versionThis work was supported in part by the National Research Foundation (NRF) of Singapore, Rolls-Royce Singapore Pte. Ltd., and in part by Nanyang Technological University, Singapore
Investigating the effects of mental healthcare accessibility on mental health outcomes in rural areas
No full textIncreasing rural-urban disparities in mental health outcomes have highlighted a critical problem at the root of mental health provision and accessibility in rural areas. In our study, we utilise the difference-in difference (DID) model and explore the impact of targeted mental health interventions on crude suicide rates in two neighbouring US rural states – Arkansas and Tennessee, from 2005 to 2018. Specifically, we assess the effects of the Arkansas Private Option, a Medicaid expansion program introduced in 2013. Our results indicate a significant reduction in crude suicide rates, with Arkansas experiencing a decrease by an average of 27.09 per 100,000 individuals compared to Tennessee after the intervention. We conclude with suggestions for further research on county-level analysis on the localised impacts of mental health intervention for different socio-economic groups, which could contribute to more equitable outcomes in mental health across diverse populations and settings.Bachelor's degre
Optimising self-amplifying RNA (saRNA) vaccines: leveraging EEEV replication mechanisms for enhanced mRNA therapy
No full textRNA therapeutics has emerged as a transformative modality capable of targeting virtually any gene, and self amplifying RNA (saRNA) platforms offer the promise of potent protein expression at dramatically reduced doses. Central to alphaviral saRNA function are the untranslated regions (UTRs), which contain core promoter elements for both minus and plus strand synthesis, cis acting replication signals, and regulatory motifs that modulate translation, innate immune evasion, and RNA stability. In this study, we focused on the 3′UTR of an Eastern equine encephalitis virus (EEEV) replicon to identify sequence segments that influence RNA replication and translational efficiency. We first performed multiple sequence alignment across New World alphaviruses and predicted secondary structures to map conserved and variable regions. Next, we engineered a series of 50 nucleotide deletion constructs, retaining the key conserved element immediately upstream of the poly(A) tail, and evaluated their performance in HEK293T cells using dual luciferase assays. Among all variants, the WTΔ50bp construct exhibited nearly two fold higher reporter activity than the full length 3′UTR, indicating that this region could possibly repress expression. These findings pinpoint a target for rational optimization of the EEEV 3′UTR and lay the groundwork for next generation saRNA vaccines with enhanced potency and lower-dose requirements.Bachelor's degre
The confluence: an open innovation ecosystem for scalable and evolutionary educational mobile chatbot
No full textLarge Language Models (LLMs) are large deep learning models with powerful capabilities such as text summary, code generation, sentiment analysis, etc. In schools with heavily imbalanced student-to-professor ratios, the introduction of LLMs in recent years serves as a promising solution for providing personalised learning support. However, implementing LLMs as educational support tools requires these LLMs to be further fine-tuned with institution-specific materials to produce responses with higher accuracy and reliability. This project introduces an Open Innovation Ecosystem designed to streamline the development and deployment of fine-tuned educational LLMs. The ecosystem provides LLM developers with a simplified integration flow, allowing them to connect their back-end endpoints directly to a pre-built infrastructure that handles front-end interfaces, middleware configuration, and database management. By abstracting these technical components, the ecosystem significantly reduces development cycles and enables developers to focus exclusively on optimising LLM outputs for educational contexts. Furthermore, the deployed LLMs are aggregated on a single mobile application as part of the ecosystem, serving as a centralised platform for students to access all relevant chatbots. This integration addresses the fragmentation issues of current implementations, where standalone chatbot applications create disconnected learning experiences. This project not only accelerates the development of educational LLMs but also greatly enhances the student learning experience with an all-in-one platform that brings together diverse LLM tools, establishing a scalable framework that serves as a blueprint for integrating LLMs within educational institutions.Bachelor's degre