Face Classification Based on Multi-task Gaussian Process Regression and Chinese Medicine Five Element System

Abstract

TCM(Traditional Chinese Medicine) physical classi-fication system based on “Five Elements” is the foundation and core material of physical study, and it is a classification system appropriate for a group's physical characteristics. Obtaining appropriate physical classification can aid in disease diagnosis efficiency. We believe that the original problem cannot be simply described as five separate tasks, like independent score inference using five different models for each element category. So we propose an approach based on the Insightface algorithm and Multi-Task Gaussian Process Regression (MTGPR) model to classify faces. MTGPR is a model that attempts to learn inter-task dependencies based solely on the task identities and the observed data for each task. It uses a parameterized covariance function over the input features x to develop a “free-form” task-similarity matrix. In MTGPR model, this is achieved by having a common covariance function over the features $x$ of the input observations. The experimental results show that our proposed method has improved results compared to the traditional Resnet-based classification method.