Learning Arbitrary Complex Matrices by Interlacing Amplitude and Phase Masks with Fixed Unitary Operations

Abstract

Programmable photonic integrated circuits represent an emerging technology that amalgamates photonics and electronics, paving the way for light-based information processing at high speeds and low power consumption. Considering their wide range of applications as one of the most fundamental mathematical operations there has been a particular interest in programmable photonic circuits that perform matrix-vector multiplication. In this regard, there has been great interest in developing novel circuit architectures for performing matrix operations that are compatible with the existing photonic integrated circuit technology which can thus be reliably implemented. Recently, it has been shown that discrete linear unitary operations can be parameterized through diagonal phase parameters interlaced with a fixed operator that enables efficient photonic realization of unitary operations by cascading phase shifter arrays interlaced with a multiport component. Here, we show that such a decomposition is only a special case of a much broader class of factorizations that allow for parametrizing arbitrary complex matrices in terms of diagonal matrices alternating with a fixed unitary matrix. Thus, we introduce a novel architecture for physically implementing discrete linear operations. The proposed architecture is built on representing an $N\times N$ matrix operator in terms of $N+1$ amplitude-and-phase modulation layers interlaced with a fixed unitary layer that could be implemented via a coupled waveguide array. The proposed architecture enables the development of novel families of programmable photonic circuits for on-chip analog information processing.

References in corpus (23)

• Deep Learning with Coherent Nanophotonic Circuits
• Photonics for artificial intelligence and neuromorphic computing
• Extreme Broadband Transparent Optical Phase Change Materials for High-Performance Nonvolatile Photonics
• Photonic quantum information processing: a concise review
• Quantum transport simulations in a programmable nanophotonic processor
• Self-Configuring Universal Linear Optical Component
• Frequency-encoded photonic qubits for scalable quantum information processing
• Implementation of Quantum and Classical Discrete Fractional Fourier Transforms
• Robust architecture for programmable universal unitaries
• Direct dialling of Haar random unitary matrices
• Three-dimensional entanglement on a silicon chip
• Programmable frequency-bin quantum states in a nano-engineered silicon device
• Dual-path source engineering in integrated quantum optics
• Photonic convolutional neural networks using integrated diffractive optics
• Addressing the programming challenges of practical interferometric mesh based optical processors
• All-optical frequency processor for networking applications
• Heavy tails and pruning in programmable photonic circuits
• Auto-calibrating Universal Programmable Photonic Circuits: Hardware Error-Correction and Defect Resilience
• The Goldilocks Principle of Learning Unitaries by Interlacing Fixed Operators with Programmable Phase Shifters on a Photonic Chip
• Symmetric silicon microring resonator optical crossbar array for accelerated inference and training in deep learning
• Waveguide lattice based architecture for multichannel optical transformations
• Lower-depth programmable linear optical processors
• Integrated Photonic Fractional Convolution Accelerator

Cited by in corpus (4)

• The Goldilocks Principle of Learning Unitaries by Interlacing Fixed Operators with Programmable Phase Shifters on a Photonic Chip
• Photonic Matrix Multiplier Makes a Direction-Finding Sensor
• Programmable Space-Frequency Linear Transformations in Photonic Interlacing Architectures
• Embedding Matrices in Programmable Photonic Networks with Flexible Depth and Width