报告人：陈靖，中科院物理研究所，博士
报告时间：2017.12.11（周一）下午3点
报告地点：LE523
摘要:Machine learning has been developing fast and achieved great success in different fields. Physicists also try to introduce the machine learning to physics. From simulating wave function, phase classification to DFT. On the one hand, we hope that the experience of algorithm and hardware will benefit the computational physics. As a concrete example, we proved the equivalence between tensor network and restricted Boltzmann Machine in machine learning. On the other hand, we show that the concept of quantum entanglement will help to qualify the power of different neural network structures.
References:
On the Equivalence of Restricted Boltzmann Machines and Tensor Network States. arXiv:1701.04831 ，Jing Chen, Song Cheng, Haidong Xie, Lei Wang, Tao Xiang
报告人简介：
Graduate at Physics from East China Normal University (2010), Master’s and DSc. at Physics from Institute of physics, CAS (2017). He will move to Flatiron Institute of Simons Foundation, New York, USA as a postdoctor.During the graduate study, Dr.Chen focus on developing DMRG and tensor network methods and applications on classical models and quantum problems. By high-order tensor network renormalization group, he and his coauthors benchmarked the 3D Ising critical point. They also calculate ground state for Anti-ferromagnetic Heisenberg on Kagome lattice by PESS methods. In the last year they had shown the inner connections between some neural network with tensor networks and proved the restricted Boltzmann machines belong to a subset of tensor network family.
