July, 2024: AltumView is a global leader in developing privacy-preserving AI solutions for healthcare. In our latest paper below, we propose the first method that applies the low-rank approximation technique in large language models (LLMs) to skeleton-based human action recognition, which can significantly reduce the complexity of the transformer architecture with negligible performance degradation or even better performance after retraining.
S. Oraki, H. Zhuang, and J. Liang, “LORTSAR: Low-Rank Transformer for Skeleton-based Action Recognition,” submitted, Jul. 2024.
Categories: Blog
Jie Liang
Dr. Jie Liang (PhD) is a Professor at Simon Fraser University, a Fellow of Canadian Academy of Engineering, and President/CEO of AltumView, which develops the Sentinare activity sensor for healthcare.