The Asian Conference on Machine Learning (ACML) 2024, hosted in Hanoi, Vietnam, is a premier forum showcasing the growing influence of the Asian ML community on the global AI landscape. Renowned for its rigorous peer-review process, ACML features high-quality research with a strong emphasis on Asia’s dynamic and rapidly evolving ML ecosystem. This year’s conference highlights cutting-edge innovations, with keynote speakers from Google DeepMind, Caltech, and the University of Illinois Urbana-Champaign (UIUC) addressing topics like large language models, neurosymbolic AI for safety-critical control, and AI-driven superhuman reasoning, cementing ACML 2024 as a key event for advancing machine learning.
Duy Nguyen from IML presents a new paper on leveraging prompt learning for few-shot adaptation in large vision-language models. The proposed algorithm employs an unbalanced optimal transport mechanism to build a bridge representation between visual features and text embeddings, effectively handling imbalanced settings with abstract text descriptions generated by GPT-3. This work is a collaboration between DFKI, the Technical University of Darmstadt, the University of Pennsylvania, and other institutions.
In addition to presenting his paper, Duy Nguyen also serves as a session chair at ACML 2024, leading discussions on multi-modal learning, with oral presentations covering topics such as diffusion models and dynamic graph learning. Furthermore, the conference provided an opportunity to establish a new connection with researchers from UIUC, setting the stage for future collaborations on foundational models for molecules, AI for science, and healthcare, which promise to drive impactful research forward.
Duy Nguyen from IML presenting his work at ACML 2024
Duy Nguyen leading a discussion on multi-modal learning at ACML 2024
