Impressive start on the first day of ACML 2024: the theory of AI superiority and a new perspective on machine learning through optimal transportation

The Asian Conference on Machine Learning (ACML) is an annual event that brings together leading experts in the field of machine learning. Held for the 16th time in Hanoi, the conference focuses on advances in both theoretical research and practical applications of machine learning, from deep learning models to optimization methods. It serves as an opportunity to exchange knowledge, connect the research community, and foster collaboration to advance machine learning in Asia and worldwide.

As a platinum sponsor and organizer of the 16th Asian Conference on Machine Learning, VinUni is proud to be a leading education and research center in Vietnam. Established with the mission of developing high-quality human resources in science, technology, and innovation, VinUni has been contributing to the creation of a groundbreaking academic and research environment. Supporting the 16th Asian Conference on Machine Learning reaffirms the university’s commitment to advancing cutting-edge engineering and technology development in the region and globally.

Let’s take a look at some of the highlights from the first day of the conference.

•  𝙋𝙡𝙚𝙣𝙖𝙧𝙮 𝙎𝙚𝙨𝙨𝙞𝙤𝙣 #1: “AI Superhuman Reasoning for Math and Beyond” by Dr. Thang Luong (Google DeepMind). 

Dr. Thang Luong began in Neural Machine Translation in 2014, advancing to reasoning-based systems like ChatGPT in 2022 and now math-solving AI nearing gold medal-level at the International Mathematics Olympiad. Combining intuitive System 1 reasoning with structured System 2 logic, it generates proofs with human-like precision. He aims for AI to surpass human abilities and predicts it will win a Fields Medal by 2030.

•  𝙋𝙡𝙚𝙣𝙖𝙧𝙮 𝙎𝙚𝙨𝙨𝙞𝙤𝙣 #2: “Learning as Distribution Matching: A Perspective through Optimal Transport” by Prof. Dinh Phung (Monash University). 

Prof. Dinh Phung highlighted his team’s groundbreaking advancements in Machine Learning, starting with the history of Optimal Transport Theory from Napoleonic-era France. This theory, originally about efficiently distributing goods while minimizing costs, underpins key machine learning problems like mapping probability distributions. His talk covered themes like robustness, hierarchical understanding, cross-domain transfer learning, and discovering new concepts.

In the afternoon, 06 oral presentation sessions were held simultaneously in different rooms at VinUniversity. Each session covered various topics, allowing participants to engage deeply with cutting-edge research and ideas in their respective fields.