论文部分内容阅读
年龄分类方法和年龄回归方法是年龄识别任务的主要方法.这两种方法各自具有其自身的优越性.例如:年龄分类方法能够灵活利用机器学习中的区分式模型,而年龄回归方法的主要优势是能够捕捉不同年龄之间的联系.为了能同时利用分类方法和回归方法的优势,本文提出了一种混合分类/回归模型(混合模型)用于用户年龄识别.具体而言,我们首先基于长短时记忆(long short-term memory,LSTM)模型分别构建年龄分类模型和回归模型用于年龄识别;在此基础上,将年龄分类结果与年龄回归结果进行线性融合作为年龄识别的最终结果.实验结果表明本文提出的混合模型能够有效提升年龄识别任务的性能.
Age classification methods and age regression methods are the main methods of age recognition tasks, each of which has its own advantages, for example: age classification method can flexibly use the discriminative model in machine learning, and the main advantages of age regression method Is able to capture the relationship between different ages.In order to make use of the advantages of both classification and regression methods at the same time, this paper presents a hybrid classification / regression model (hybrid model) for user age identification.Firstly, we first based on the length Time long-term memory (LSTM) model was used to construct the age classification model and the regression model for age identification respectively. On the basis of this, linear fusion of age classification results and age regression results was used as the final result of age identification. It shows that the hybrid model proposed in this paper can effectively improve the performance of age recognition tasks.