Neuromorphic photonic networks using silicon photonic weight banks

Abstract

Photonic systems for high-performance information processing have attracted renewed interest. Neuromorphic silicon photonics has the potential to integrate processing functions that vastly exceed the capabilities of electronics. We report first observations of a recurrent silicon photonic neural network, in which connections are configured by microring weight banks. A mathematical isomorphism between the silicon photonic circuit and a continuous neural network model is demonstrated through dynamical bifurcation analysis. Exploiting this isomorphism, a simulated 24-node silicon photonic neural network is programmed using “neural compiler” to solve a differential system emulation task. A 294-fold acceleration against a conventional benchmark is predicted. We also propose and derive power consumption analysis for modulator-class neurons that, as opposed to laser-class neurons, are compatible with silicon photonic platforms. At increased scale, Neuromorphic silicon photonics could access new regimes of ultrafast information processing for radio, control, and scientific computing.

Fundamental aspects of noise in analog-hardware neural networks

“…The variance for the 2nd layer is Var(y t 2,n ) = α 2 Var(y t 1 ). Equations (12) and (13) are recursively defined in function of the network's depth. From these relationships it becomes clear that the variance of hidden layers does not contain uncorrelated noise-contributions originating from the previous hidden layers.…”

A Winograd-Based Integrated Photonics Accelerator for Convolutional Neural Networks

“…For this scheme, according to the bit resolution, the estimated dynamic range is 20dB. The speed of the optical part of the accelerator, without considering the I/O interface, according to [29] is given by the total number of the MRR and their pitch. Photodetection and phase cross-talk are expected to be the main sources of error in the proposed scheme.…”

ITO-based electro-absorption modulator for photonic neural activation function

“…The activation function was replaced with a custom designed non linear transfer function resulting from our ab-initio ITO based electro-absorption modulator model. The weighting mechanism relies on microring resonator banks as proposed by Tait et al 13,25 , which can potentially support more than 100 channels 13 . In this case, the weights were bound between minus one and one to simulate input optical weighting by rings in a push-pull configuration.…”

“…More in details, interesting experimental and numerical all-optical nonlinear module based on saturable and reverse absorption 15,16 , graphene excitable lasers 17,18 , twosection distributed-feedback (DFB) lasers 19 , quantum dots 20 , disks lasers 21,22 , induced transparency in quantum assembly 15 have recently been reported and showed promising results in terms of efficiency and throughput for different kinds of neural network and applications, ranging from convolutional neural network, spiking neural network and reservoir computing. Although, a more straightforward implementation is currently attained by exploiting electro-optic tuned nonlinear materials 23,24 or absorptive modulator directly connected to a photodiode, as shown in 14,[25][26][27][28] . In this case, the photogenerated current, proportional to the detected optical power at the weighted addition, alters the voltage drop on the active material, thus changing its carrier concentration and consequently the effective modal index of the propagating waveguide mode.…”