课题针对传统机器学习方法在非结构化环境下存在的局限和目前新人工智能发展的瓶颈，拟借鉴生物大脑的信息感知机制与认知机理，发展感知-认知-强化为一体的类脑计算新理论与新方法，围绕模型、算法与计算三部分关键问题展开研究。具体包括：1）目标信息类脑感知模式分析；2）类脑认知机器学习模型构建；3）多模态深度强化学习建模与优化算法；4）面向遥感解译应用的高效认知计算与实现。期望通过上述研究，在非结构化环境下提升机器学习模型的环境感知能力、认知与交互能力，以及学习算法在执行复杂任务时的有效性与普适性。在理论研究的基础上，研制轻量级的实用类脑认知机器学习模型，在高分辨遥感数据解译任务中进行验证。本项目的研究背景明确，原创性高，研究成果对于促进人工智能基础与应用研究具有重要意义与价值。

Aiming at the limitation of traditional machine learning methods and the bottleneck of the development of new artificial intelligence, we refer to the perception and cognition mechanism of biological brain to develop new theories and new methods of brain-like computation which gathers perception, cognition and reinforcement as a whole. And research on three key issues of existing machine learning, which are data, model and algorithm. The contents of this project include: 1) Analysis of brain-like perception model of object information; 2) Establishing brain-inspired cognitive machine learning model; 3) Modeling and optimizing the multimodal deep reinforcement learning; 4) Efficient cognitive calculation and implementation for Interpretation of Remote Sensing. It is expected that under the non-structure environment, the environmental perception and the environmental cognition and interaction ability of machine can be improved. Also, the universality of learning algorithms while performing complex learning tasks can be enhanced. Moreover, based on the theoretical work, we will developing a lightweight and practical brain-like cognitive machine learning model, and validate it in high resolution remote sensing interpretation task. This project has a clear research background, high level of originality, and the results will be valued and have a far-reaching impact for promoting the basic and application research on Artificial Intelligence.