2018.09.28 Dynamical Principles of Neural Information Processing

2019-07-07 00:57:41 0



    目: Dynamical Principles of Neural Information Processing





主持人:汤超 教授



  Recently, we have witnessed the great success of deep neural networks (DNNs) in many applications. Compared to DNNs, which mainly capture the hierarchical feedforward architecture of the visual pathway, the real neural system has much richer structures and possess much more powerful computational capacities. These different structures and their associated functions lay the foundation for us to develop new advanced brain-inspired computational algorithms. In this talk, I will briefly review some of these key differences, with a focus on the dynamical aspects of neural information processing.



  吴思,北京大学信息科学技术学院教授、博士生导师,麦戈文脑科学所研究员。研究方向为计算神经科学和类脑计算,通过和实验神经科学家紧密合作,以数学理论和计算机仿真来构建神经系统的计算模型,解析神经系统处理信息的基本原理,并在此基础上发展类脑智能算法。目前担任计算神经科学领域学术期刊Frontiers in Computational Neuroscience共同主编。已发表论文上百篇,包括神经科学的顶级杂志NeuronNature NeurosciencePNASJ. Neurosci.等,以及人工智能顶级国际会议NIPS等。