DR-NTU (Digital Repository of NTU)
Not a member yet
116018 research outputs found
Sort by
Sea level change in Asia
No full textThe consequences of sea level rise present some of the most formidable challenges humanity faces. Global mean sea level rose faster in the 20th century than any prior century in the last three millennia. It is also virtually certain that the global mean sea level will continue to rise up to 2100. However, sea level changes exhibit substantial spatial variability, indicating the need for region-to-local scale studies. Understanding and communicating the contributing factors of sea level change requires observations and process models that span both global and local scales involving interdisciplinary research teams in Earth science. This chapter synthesizes contemporary past and future sea level changes and their impacts with a particular focus on the Asian region. The chapter also covers advances since the Intergovernmental Panel on Climate Change Sixth Assessment Report (IPCC AR6) published in 2021.
Anthropogenic climate change has contributed substantially to global mean sea level rise since 1900; however, sea level changes are not uniform globally and can vary up to ±20% regionally. The variations in sea level in deeper oceans are primarily influenced by steric components, while shallower oceans and continental shelves are mainly affected by manometric sea level components. Asian coastal areas are highly vulnerable to 21st-century sea level rise, with Japan’s east coast projected to experience the highest magnitude in the region. Extreme sea levels are likely to increase in the region, with 9–10 m of worst-case extreme sea level rise possible on the China coast. Appropriate adaptation and mitigation strategies to reduce the impact of sea level rise should start now
Deep reinforcement learning for table tennis
No full textDeep Reinforcement Learning (DRL) is a powerful branch of machine learning that
combines deep learning with reinforcement learning (RL) to enable intelligent agents
to learn complex behaviors. DRL has achieved remarkable success in areas such as
robotics, game playing (e.g., AlphaGo, OpenAI Five), autonomous vehicles, and
finance.
This report compares various configurations of DRL algorithms to train agents in
games. This report also continues with the work by Professor Seah Hock Soon and
his team, who built a novel table tennis game in the Unity environment and
successfully trained agents of different proportions to perform forehand shots using
DRL.
We attempted to teach an agent more complicated tasks. Firstly, we experimented
with backhand shots before attempting to teach an agent to perform both forehand
and backhand shots appropriately. Through these attempts, we discovered more
methods to improve results and training with DRL.
The key contribution of this report is the discovery of methods to improve training
speed and effectiveness DRL training, especially when rewards are sparse, like in
the table tennis environment, which can be valuable when DRL is applied to other
areas.Bachelor's degre
Support-free iridium hydroxide for high-efficiency proton-exchange membrane water electrolysis
No full textThe large-scale implementation of proton-exchange membrane water electrolyzers relies on high-performance membrane-electrode assemblies that use minimal iridium (Ir). In this study, we present a support-free Ir catalyst developed through a metal-oxide-based molecular self-assembly strategy. The unique self-assembly of densely isolated single IrO6H8 octahedra leads to the formation of μm-sized hierarchically porous Ir hydroxide particles. The support-free Ir catalyst exhibits a high turnover frequency of 5.31 s⁻¹ at 1.52 V in the membrane-electrode assembly. In the corresponding proton-exchange membrane water electrolyzer, notable performance with a cell voltage of less than 1.75 V at 4.0 A cm⁻² (Ir loading of 0.375 mg cm⁻²) is achieved. This metal-oxide-based molecular self-assembly strategy may provide a general approach for the development of advanced support-free catalysts for high-performance membrane-electrode assemblies.Agency for Science, Technology and Research (A*STAR)National Research Foundation (NRF)Published versionThis work is supported by the National Natural Science Foundation of China (No. 22409174), The National Key Research and Development Program of China (No. 2024YFA1509200, No. 2024YFA1509203, No. 2024YFA1509204), and Startup Foundation for Hundred-Talent Program of Zhejiang University to Y.C. This work is also supported by the Agency for Science, Technology and Research (A*STAR) MTC Individual Research Grants (IRG) M22K2c0078 and Singapore National Research Foundation under its Campus for Research Excellence and Technological Enterprise (CREATE) programme to Z.J.X
InsAIghts: dynamic dashboards powered by generative AI
No full textIn an era where data is growing at an unprecedented rate, the ability to make sense of large, complex datasets and transform them into actionable insights is more critical than ever. This report focuses on simplifying data visualisation and analysis to empower users to harness information and make better decisions.
This report provides an overview of the development process of InsAIghts, an application to create dashboards by leveraging generative AI designed to make data analysis more accessible and efficient. This report explores the research conducted, the approach adopted and the technical implementation of the application. Furthermore, it discusses the current limitations and outlines potential future enhancements and additional features aimed at improving the functionality and usability of InsAIghts.Bachelor's degre
Brain-controlled pacman: a hand movement training game
No full textAdvancements in brain-computer interface (BCI) technologies have enabled the decoding of motor intentions from non-invasive brain signals. However, continuous decoding of fine motor movements, particularly hand grasping, remains underexplored. This project presents Brain-Controlled Pacman, a gamified training platform that integrates real-time electroencephalography (EEG)-based decoding with interactive gameplay to
facilitate hand motor control.
The system leverages a pre-trained deep learning model to continuously predict fingerjoint bending angles from EEG signals captured during grasping tasks. These predictions are mapped onto the in-game actions of a Pacman character in a custom-built game, Attack on Ghost Gang, where players must collect fruit and defeat enemies using their brain-controlled hand motions.
Through this platform, the project investigates the feasibility and responsiveness of EEG-driven control in a dynamic environment, while analyzing how repeated task execution affects motor performance. The game design incorporates progressive difficulty levels to examine grasp intensity under increasing task demands. By combining BCI, deep learning, and game-based interaction, this work contributes to the growing field
of neurorehabilitation by offering an engaging tool for training and evaluating hand motor functionBachelor's degre
Pakar-an intelligent mobile app for car park recommendation (front-end)
No full textThe study of innovative data visualisation techniques to visualise car park information dataBachelor's degre
My dog Is stupid
No full textMy Dog Is Stupid is a 2D solo animated film written, designed and animated by Samantha Goh. The narrative follows a young fisherman and his dog in a stranded boat, in which the fisherman tries to signal
to a nearby stranger for help. Throughout the film, his companion dog creates problems for him. This film
was heavily inspired by observing the funny relationships of animal companions in Western animation,
which are mostly portrayed as a comedic sidekick or an emotional anchor that builds the narration
forward.Bachelor's degre
Conservapedia: problems, principles, mindset and implications for a polarized world
No full textPurpose: In this article, I present an initial examination of Conservapedia; namely, the problems it identified in the Wikipedia project that made a split appear necessary and the principles it claims to follow. I then argue that Conservapedia is characterized by a “law-and-order” mindset. Finally, implications for the continued existence of Conservapedia in a polarized world are presented. Design/methodology/approach: A content analysis of key Conservapedia documents was conducted. Findings: The founders of Conservapedia took issue with Wikipedia over its supposed intolerance and inconsistency of thought. They developed a set of principles that attempted to reconcile open-mindedness with efficiency and an extreme point of view on certain subjects. Nevertheless, Conservapedia failed to produce a vibrant community, and its function today is more of a database of alt-right dogma controlled by a core group of supporters. Originality/value: There has been little scholarly attention paid to the various offshoots of Wikipedia, including Conservapedia. This is unfortunate. These alternative wiki encyclopedias represent knowledge universes of their own and in an increasingly polarized world they are important phenomena to understand.Submitted/Accepted versio
Font matters: Times New Roman increases risk aversion
No full textThis study examines whether font type influences individuals’ financial decision-making in the domains of risk and time. Drawing on principles from cognitive psychology and behavioural economics and motivated by evidence that minor typographic adjustments significantly impact readers' information reception, an online experiment was conducted with 168 participants, who were randomly assigned to either Times New Roman (a serif font) or Comic Sans (a sans-serif font). Risk preferences were measured using a lottery-choice task, while time preferences were assessed via a delay-discounting method. Results indicate that participants exposed to Times New Roman exhibited significantly greater risk aversion at a 5% level than those exposed to Comic Sans, while time preferences remained unaffected. Although participants reported distinct perceptions and emotions associated with each font, these subjective assessments did not fully account for the observed statistical behavioural differences. The findings suggest that typography may influence decision-making through brain activation and other potential mechanisms. This insight underscores the importance of design considerations in decision architecture and invites further research on the cognitive pathways through which font types can nudge economic behaviour in decision-making.Bachelor's degre
Protein-reconstructed artificial epidermis for transdermal permeation testing
No full textTransdermal permeation testing is important in assessing therapeutic molecules in applications such as cosmetics and pharmaceuticals. Progress in these fields has been hindered by the lack of inexpensive skin models that closely mimic human skin, especially in terms of transdermal permeation properties. The thesis proposes to form a protein-reconstructed artificial epidermis. An artificial epidermis, having similar transdermal permeation properties as human skin, could be fabricated by reconstructing proteins similar with human skin.
After comparing common existing protein-based biological materials, the morphological and compositional affinities of the silk fibroin with the cornified cell envelope in human skin, which contributes to the skin barrier function, assure its suitability for serving as the ingredient. To start with, a fabrication protocol from raw silk to the silk fibroin-based film with uniform thickness and controllable surface-free energy is introduced. Thin films with uniform thickness and controllable surface free energy to that of human skin are fabricated by blade coating concentrated PEGylated silk fibroin solution on different molds. The surface free energy of these films can be tailored to match that of the mold used in the blade coating process, allowing the replication of human skin's SFE, such as forearm skin, by selecting specific materials like aluminum foil.
Then, multiple silk fibroin thin films are hot-pressed together into one thick film to serve as the protein-reconstructed artificial epidermis. The hot pressing, conducted under 4 MPa, requires a temperature higher than 120 °C, ensuring that these films be seamlessly integrated. The increase of the operation temperature in hot pressing gives a rise in the ratio of β-sheet/β-turn and at the expense of random coils, leading to greater crystallinity with no significant chemical change like degradation or oxidation. Experiments, characterizations, and simulations results show that both temperature and pressure play an important role.
The study also developed a protein and lipid-reconstructed artificial epidermis. Both types of artificial epidermis demonstrated transdermal permeation profiles closely matching human skin and surpassed existing models, Episkin and Strat-M, in permeability speed test with different compounds. The penetration mechanism for small compounds was studied, showing that molecules primarily permeate through protein via random coils rather than denser structures, like β-sheets/β-turns or α-helices. Additionally, compounds with low molecular weight and high aqueous solubility exhibited better permeation rates on these artificial epidermises, aligning with established summary of different physicochemical factors’ influence on permeability. These findings further validate the potential application of protein-reconstructed artificial epidermis for transdermal permeation testing.Doctor of Philosoph