基于区域选举稳定性理论的人工智能方法研究

项目名称： 基于区域选举稳定性理论的人工智能方法研究

项目编号： No.61473212

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 其他

项目作者： 陈亮

作者单位： 温州大学

项目金额： 67万元

中文摘要： 经过数十年的快速发展，人工智能和模式识别领域的一些基础性问题仍然具有非常大的挑战性，例如人脑具体的决策机制及其实现方式。本项目拟从社会科学中的选举制度的稳定性研究出发，针对低质量图像的人脸识别问题开展一系列原创性的研究工作，力求实现在选举制度稳定性理论方面及其在人工智能中的应用方面(低质量图像人脸识别问题)取得突破性进展。项目的难点体现在：1)如何给出在复杂因素影响下选举结果的稳定性模型和选举结果发生转变的临界条件，2)如何基于选举稳定性理论实现稳定性强的机器学习方法解决低质量人脸图像情况下的识别问题，和3)如何利用构造与区域识别法相适应，且可利用数据结构信息的深度学习模型。根据本项目的难点，我们的具体研究子目标包括1）多候选人的单一或多个获胜者情况下区域选举制度稳定性理论，2）基于区域投票理论的高精度低质量图像人脸识别方法，和3）基于深度学习和区域选举理论的人脸识别方法。

中文关键词： 人工智能；计算智能；神经网络；流形学习；模式分类

英文摘要： After decades of fast development, the fundamental problems in artificial intelligence and pattern recognition remain unsolved, for example, how exactly human brain works is still beyond our knowledge. This project is supposed to study the stability theory of regional voting of social science. Based on the stability theory, we are going to start a series of original scientific research and try our best to accomplish breakthrough level research in artificial intelligence. The main challenges in this project are: 1) How to obtain the turning point of the voting result under complex environment. 2) How to design state-of-the-art machine learning machine learning algorithms based on our stability theory for low-quality face image recognition system. 3) How to design novel deep learning neural network models based on the regional matching for low-quality face image recognition with high accuracy. To address these problems, the research of this project includes three folds: 1) The theory of regional voting for multiple candidate and multiple winning cases. 2) The state-of-the-art machine learning algorithms based on stability theory for high performance face recognition with low-quality images. 3) Deep learning algorithms based on the stability theory for face recognition.

英文关键词： Artificial intelligence;Computational intelligence;Neural networks;Manifold learning;Pattern classification