The 363th Wenlan Financial Forum

Abstract：

This presentation is based on a recent actuarial paper published in Astin Bulletin, introducing a novel network called KAN (Kolmogorov-Arnold Network) and its application in mortality modeling. KAN is a recently emerged network over the past year or two. Building upon conventional neural networks and their latest extension, CANN (Combined Actuarial Neural Net), the presentation systematically introduces the application of KANN in mortality modeling and compares it with various existing methods. The findings demonstrate that both KAN and KANN offer advantages over existing approaches in terms of model interpretability and smoothness of fitted curves.

Speaker Introduction：

Lianzeng Zhang is a professor and doctoral supervisor at the Nankai-Taikang Insurance and Actuarial Research Institute, Nankai University. In recent years, his research has focused on actuarial data science and quantitative risk management. Over the past few years, he has published papers in journals such as Insurance: Mathematics and Economics, Scandinavian Actuarial Journal, ASTIN Bulletin, Geneva Papers on Risk and Insurance—Issues and Practice, Insurance Studies, and Statistical Research. Additionally, Professor Zhang Lianzeng serves as a council member of the China Association of Actuaries, has served as the editor-in-chief of actuarial mathematics textbooks, and is responsible for compiling one of the textbooks in the field of actuarial data science. He has made significant contributions to the teaching and research of actuarial science in China and holds considerable influence within the domestic actuarial industry.