Events
Beyond scale: efficient pre-training and controllable post-training for language models
ISTD PhD Oral Defence Seminar by Zeng Guangtao – Language models are foundational to modern artificial intelligence, but their development is often constrained by challenges in efficiency, controllability, and reasoning. In this thesis, we aim to address these limitations by introducing advanced paradigms at both the pre-training and post-training stages.
The poet as experiencer – poetic consciousness and nonhuman intelligence
HASS Talk by Adam Staley Groves – Presenting his recent monograph, The Poet as Experiencer: Wallace Stevens and Nonhuman Intelligence (Punctum Books, 2025) which proposes modernist poet Wallace Stevens as a perceiver of nonhuman intelligence (NHI).
Context-aware perception in adverse conditions
ISTD PhD Oral Defence Seminar by Tan Yu Xiang – Adverse conditions such as rain or murky water environment, significantly impacts various perception tasks. In rain conditions, the images captured are easily corrupted by both raindrops on the lenses and lens flare. Meanwhile in turbid underwater conditions, the murkiness reduces the contrast and saturation of the image. To tackle these problems, we utilise contextual information to improve robustness of perception algorithms.
Analysis and synthesis of audio with AI: from neurological disease to accented speech and music
ISTD PhD Oral Defence Seminar by Jan Melechovsky – In the modern era, new technology is opening opportunities to help various groups of people around the world. In this thesis, deep learning and audio processing is utilized to target the needs of and develop specific applications for patients with progressive neurological diseases, speakers of non-native English accents, and amateur and leisure musicians and music enjoyers.
Beyond optimal methods for minimax optimisation
ESD Seminar by Chengchang Liu – This talk will introduce several novel methods that achieve even faster convergence rates or better computational complexities compared to various optimal methods by effectively incorporating curvature information and leveraging the min-max structure.
Time-sensitive AI systems for physical agents: from DNN perception to LLM planning
ESD Seminar by Neiwen Ling – Research on developing time-sensitive AI systems, focusing on Deep Neural Network (DNN)-based perception and LLM-driven planning will be presented in this talk.
Linux under the hood: powering everything from your laptop to the cloud
ISTD COIL Seminar by Norman Hsu Chen-Wei – A brief, engaging introduction to Linux as the powerful, often unseen operating system behind much of the digital world.
Advancing signal processing with modulo sampling: theory, algorithms, and applications
ISTD PhD Oral Defence Seminar by Qi Zhang – Analog-to-digital converters (ADCs) are crucial in signal processing but face challenges when handling high-dynamic-range signals. In radar systems, the coexistence of strong and weak targets can lead to significant information loss due to ADC limitations.
Airspace and air traffic management in the age of uncrewed aerial systems
ESD Research Seminar by Max Li – Recent projects related to weather-dependent service reliability of drone package deliver operations, UAS operations optimisation for hazardous environment navigation, as well as airspace congestion management for larger, air taxi-type vehicles will be discussed in this talk.
Neural network-defined physical layer: a new paradigm for software radio in the IoT era
ISTD PhD Oral Defence Seminar by Wang Jiazhao – The increase of the Internet of Things (IoT) has created a complex and heterogeneous wireless ecosystem, demanding IoT gateways that are both flexible and efficient. While Software Defined Radio (SDR) provides the necessary hardware adaptability, its potential is frequently undermined by significant software implementation challenges, including a lack of portability across platforms, prohibitive design complexity for advanced algorithms, and poor computational efficiency. This thesis posits that these persistent bottlenecks can be overcome by a paradigm shift in physical layer (PHY) design: reframing core communication functionalities as learnable, interpretable neural network (NN) models.