The 2nd Summer School on Mathematical Stat and ML (viasm.edu.vn) was a huge success. This is part of our mission to help Vietnam become a regional hub in Statistics and Data Science. Looking forward to the 2025 edition. Let me know if you want to get involved
I will be at Bayes on the Beach next week :D. Very excited to present my work at one of my favourite conferences ;) . Registration is still open.
research.qut.edu.au/qutcds/bayes-o…
During the Vietnam War, one of the most dangerous jobs was undertaken by a select few known as "tunnel rats."
These unsung heroes were American, Australian, and New Zealand soldiers specially trained as combat engineers, who crawled through Viet Cong underground tunnels to perform perilous covert search and destroy missions.
Tunnel rats gently prodded for armed mines in order to disarm them — and prayed that they survived with both their legs intact. Most men were volunteers and tended to be of smaller stature, making it easier for them to maneuver through the cramped subterranean spaces.
Natural gradient works efficiently in learning (Amari, 1998), but computing it remains a challenging problem.
Our new work arxiv.org/abs/2312.09633 completely addresses this challenge.
My paper arxiv.org/abs/2303.13930 (with P. Tseng and @robertjk59) proposes a particle-based approach that relaxes the conjugate priors requirement in MFVB. The theoretical basis is established by leveraging the connection between Wasserstein gradient flows and Langevin diffusions.
Normality of the summary statistics has been a strict assumption in synthetic likelihood. Our paper arxiv.org/abs/2305.14746 fulfills this requirement by using OT to construct a Gaussianization transformation. Joint work with PhD student N. Nguyen, @chris_drovandi and D. Nott