Espinosa-Ortega T, Liew T C H
Division of Physics and Applied Physics, Nanyang Technological University, Singapore 637371, Singapore.
Phys Rev Lett. 2015 Mar 20;114(11):118101. doi: 10.1103/PhysRevLett.114.118101. Epub 2015 Mar 18.
A general scheme to realize a perceptron for hardware neural networks is presented, where multiple interconnections are achieved by a superposition of Schrödinger waves. Spatially patterned potentials process information by coupling different points of reciprocal space. The necessary potential shape is obtained from the Hebbian learning rule, either through exact calculation or construction from a superposition of known optical inputs. This allows implementation in a wide range of compact optical systems, including (1) any nonlinear optical system, (2) optical systems patterned by optical lithography, and (3) exciton-polariton systems with phonon or nuclear spin interactions.
提出了一种用于硬件神经网络的感知器实现方案，其中通过薛定谔波的叠加实现多个互连。空间图案化势通过耦合倒易空间的不同点来处理信息。必要的势形状可从赫布学习规则获得，可通过精确计算或由已知光学输入的叠加构建得到。这使得能够在多种紧凑型光学系统中实现，包括(1)任何非线性光学系统，(2)通过光刻图案化的光学系统，以及(3)具有声子或核自旋相互作用的激子极化激元系统。
