Arbitrary Linear Transformations Research Articles

Perfect linear optics using silicon photonics

Recently there has been growing interest in using photonics to perform the linear algebra operations of neuromorphic and quantum computing applications, aiming at harnessing silicon photonics’ (SiPho) high-speed and energy-efficiency credentials. Accurately mapping, however, a matrix into optics remains challenging, since state-of-the-art optical architectures are sensitive to fabrication imperfections. This leads to reduced fidelity that degrades as the insertion losses of the optical matrix nodes or the matrix dimensions increase. In this work, we present the experimental deployment of a 4 × 4 coherent crossbar (Xbar) as a silicon chip and validate experimentally its theoretically predicted fidelity restoration credentials. We demonstrate the experimental implementation of 10,000 arbitrary linear transformations achieving a record-high fidelity of 99.997% ± 0.002, limited mainly by the measurement equipment. Our work represents an integrated optical circuit providing almost unity and loss-independent fidelity in the realization of arbitrary matrices, highlighting light’s credentials in resolving complex computations.

Universal linear intensity transformations using spatially incoherent diffractive processors

Under spatially coherent light, a diffractive optical network composed of structured surfaces can be designed to perform any arbitrary complex-valued linear transformation between its input and output fields-of-view (FOVs) if the total number (N) of optimizable phase-only diffractive features is ≥~2NiNo, where Ni and No refer to the number of useful pixels at the input and the output FOVs, respectively. Here we report the design of a spatially incoherent diffractive optical processor that can approximate any arbitrary linear transformation in time-averaged intensity between its input and output FOVs. Under spatially incoherent monochromatic light, the spatially varying intensity point spread function (H) of a diffractive network, corresponding to a given, arbitrarily-selected linear intensity transformation, can be written as H(m, n; m′, n′) = |h(m, n; m′, n′)|2, where h is the spatially coherent point spread function of the same diffractive network, and (m, n) and (m′, n′) define the coordinates of the output and input FOVs, respectively. Using numerical simulations and deep learning, supervised through examples of input-output profiles, we demonstrate that a spatially incoherent diffractive network can be trained to all-optically perform any arbitrary linear intensity transformation between its input and output if N ≥ ~2NiNo. We also report the design of spatially incoherent diffractive networks for linear processing of intensity information at multiple illumination wavelengths, operating simultaneously. Finally, we numerically demonstrate a diffractive network design that performs all-optical classification of handwritten digits under spatially incoherent illumination, achieving a test accuracy of >95%. Spatially incoherent diffractive networks will be broadly useful for designing all-optical visual processors that can work under natural light.

Universal Time-Dependent Control Scheme for Realizing Arbitrary Linear Bosonic Transformations.

We study the implementation of arbitrary excitation-conserving linear transformations between two sets of N stationary bosonic modes, which are connected through a photonic quantum channel. By controlling the individual couplings between the modes and the channel, an initial N-partite quantum state in register A can be released as a multiphoton wave packet and, successively, be reabsorbed in register B. Here we prove that there exists a set of control pulses that implement this transfer with arbitrarily high fidelity and, simultaneously, realize a prespecified N×N unitary transformation between the two sets of modes. Moreover, we provide a numerical algorithm for constructing these control pulses and discuss the scaling and robustness of this protocol in terms of several illustrative examples. By being purely control-based and not relying on any adaptations of the underlying hardware, the presented scheme is extremely flexible and can find widespread applications, for example, for boson-sampling experiments, multiqubit state transfer protocols, or in continuous-variable quantum computing architectures.

Polarization multiplexed diffractive computing: all-optical implementation of a group of linear transformations through a polarization-encoded diffractive network

Research on optical computing has recently attracted significant attention due to the transformative advances in machine learning. Among different approaches, diffractive optical networks composed of spatially-engineered transmissive surfaces have been demonstrated for all-optical statistical inference and performing arbitrary linear transformations using passive, free-space optical layers. Here, we introduce a polarization-multiplexed diffractive processor to all-optically perform multiple, arbitrarily-selected linear transformations through a single diffractive network trained using deep learning. In this framework, an array of pre-selected linear polarizers is positioned between trainable transmissive diffractive materials that are isotropic, and different target linear transformations (complex-valued) are uniquely assigned to different combinations of input/output polarization states. The transmission layers of this polarization-multiplexed diffractive network are trained and optimized via deep learning and error-backpropagation by using thousands of examples of the input/output fields corresponding to each one of the complex-valued linear transformations assigned to different input/output polarization combinations. Our results and analysis reveal that a single diffractive network can successfully approximate and all-optically implement a group of arbitrarily-selected target transformations with a negligible error when the number of trainable diffractive features/neurons (N) approaches N_pN_iN_o, where Ni and No represent the number of pixels at the input and output fields-of-view, respectively, and Np refers to the number of unique linear transformations assigned to different input/output polarization combinations. This polarization-multiplexed all-optical diffractive processor can find various applications in optical computing and polarization-based machine vision tasks.

Arbitrary optical wave evolution with Fourier transforms and phase masks.

A large number of applications in classical and quantum photonics require the capability of implementing arbitrary linear unitary transformations on a set of optical modes. In a seminal work by Reck et al. [Phys. Rev. Lett.73, 58 (1994)10.1103/PhysRevLett.73.58], it was shown how to build such multiport universal interferometers with a mesh of beam splitters and phase shifters, and this design became the basis for most experimental implementations in the last decades. However, the design of Reck et al. is difficult to scale up to a large number of modes, which would be required for many applications. Here we present a deterministic algorithm that can find an exact and efficient implementation of any unitary transformation, using only Fourier transforms and phase masks. Since Fourier transforms and phase masks are routinely implemented in several optical setups and they do not suffer from some of the scalability issues associated with building extensive meshes of beam splitters, we believe that our design can be useful for many applications in photonics.

All-optical synthesis of an arbitrary linear transformation using diffractive surfaces

Spatially-engineered diffractive surfaces have emerged as a powerful framework to control light-matter interactions for statistical inference and the design of task-specific optical components. Here, we report the design of diffractive surfaces to all-optically perform arbitrary complex-valued linear transformations between an input (Ni) and output (No), where Ni and No represent the number of pixels at the input and output fields-of-view (FOVs), respectively. First, we consider a single diffractive surface and use a matrix pseudoinverse-based method to determine the complex-valued transmission coefficients of the diffractive features/neurons to all-optically perform a desired/target linear transformation. In addition to this data-free design approach, we also consider a deep learning-based design method to optimize the transmission coefficients of diffractive surfaces by using examples of input/output fields corresponding to the target transformation. We compared the all-optical transformation errors and diffraction efficiencies achieved using data-free designs as well as data-driven (deep learning-based) diffractive designs to all-optically perform (i) arbitrarily-chosen complex-valued transformations including unitary, nonunitary, and noninvertible transforms, (ii) 2D discrete Fourier transformation, (iii) arbitrary 2D permutation operations, and (iv) high-pass filtered coherent imaging. Our analyses reveal that if the total number (N) of spatially-engineered diffractive features/neurons is ≥Ni × No, both design methods succeed in all-optical implementation of the target transformation, achieving negligible error. However, compared to data-free designs, deep learning-based diffractive designs are found to achieve significantly larger diffraction efficiencies for a given N and their all-optical transformations are more accurate for N < Ni × No. These conclusions are generally applicable to various optical processors that employ spatially-engineered diffractive surfaces.

Arbitrary linear transformations for photons in the frequency synthetic dimension

Arbitrary linear transformations are of crucial importance in a plethora of photonic applications spanning classical signal processing, communication systems, quantum information processing and machine learning. Here, we present a photonic architecture to achieve arbitrary linear transformations by harnessing the synthetic frequency dimension of photons. Our structure consists of dynamically modulated micro-ring resonators that implement tunable couplings between multiple frequency modes carried by a single waveguide. By inverse design of these short- and long-range couplings using automatic differentiation, we realize arbitrary scattering matrices in synthetic space between the input and output frequency modes with near-unity fidelity and favorable scaling. We show that the same physical structure can be reconfigured to implement a wide variety of manipulations including single-frequency conversion, nonreciprocal frequency translations, and unitary as well as non-unitary transformations. Our approach enables compact, scalable and reconfigurable integrated photonic architectures to achieve arbitrary linear transformations in both the classical and quantum domains using current state-of-the-art technology.

Affine-transformation invariant clustering models

We develop a cluster process which is invariant with respect to unknown affine transformations of the feature space without knowing the number of clusters in advance. Specifically, our proposed method can identify clusters invariant under (I) orthogonal transformations, (II) scaling-coordinate orthogonal transformations, and (III) arbitrary nonsingular linear transformations corresponding to models I, II, and III, respectively and represent clusters with the proposed heatmap of the similarity matrix. The proposed Metropolis-Hasting algorithm leads to an irreducible and aperiodic Markov chain, which is also efficient at identifying clusters reasonably well for various applications. Both the synthetic and real data examples show that the proposed method could be widely applied in many fields, especially for finding the number of clusters and identifying clusters of samples of interest in aerial photography and genomic data.

Swap and Rotate: Lightweight Linear Layers for SPN-based Blockciphers

In CHES 2017, Jean et al. presented a paper on “Bit-Sliding” in which the authors proposed lightweight constructions for SPN based block ciphers like AES, PRESENT and SKINNY. The main idea behind these constructions was to reduce the length of the datapath to 1 bit and to reformulate the linear layer for these ciphers so that they require fewer scan flip-flops (which have built-in multiplexer functionality and so larger in area as compared to a simple flip-flop). In this paper, we develop their idea even further in few separate directions.First, we prove that given an arbitrary linear transformation, it is always possible to construct the linear layer using merely 2 scan flip-flops. This points to an optimistic venue to follow to gain further GE reductions, yet the straightforward application of the techniques in our proof to PRESENT and GIFT leads to inefficient implementations of the linear layer, as reducing ourselves to 2 scan flip-flops setting requires thousands of clock cycles and leads to very high latency.Equipped with the well-established formalism on permutation groups, we explore whether we can reduce the number of clock cycles to a practical level, i.e. few hundreds, by adding few more pairs of scan flip flops. For PRESENT, we show that 4 (resp. 8, 12) scan flip-flops are sufficient to complete the permutation layer in 384 (resp. 256, 128) clock cycles. For GIFT, we show that 4 (resp. 8, 10) scan flip flops correspond to 320 (resp. 192, 128) clock cycles. Finally, in order to provide the best of the two worlds (i.e. circuit area and latency), we push our scan flip-flop choices even further to completely eliminate the latency incurred by the permutation layer, without compromising our stringent GE budget. We show that not only 12 scan flip flops are sufficient to execute PRESENT permutation in 64 clock cycles, but also the same scan flip flops can be used readily in a combined encryption decryption circuit. Our final design of PRESENT and GIFT beat the record of Jean et al. and Banik et al. in both latency and in circuit-size metric. We believe that the techniques presented in our work can also be used at choosing bit-sliding-friendly linear layer permutations for the future SPN-based designs.

On single-layer potentials for a class of pseudo-differential equations related to linear transformations of a symmetric $\alpha$-stable stochastic process

In this article an arbitrary invertible linear transformations of a symmetric $\alpha$-stable stochastic process in $d$-dimensional Euclidean space $\mathbb{R}^d$ are investigated. The result of such transformation is a Markov process in $\mathbb{R}^d$ whose generator is the pseudo-differential operator defined by its symbol $(-(Q\xi,\xi)^{\alpha/2})_{\xi\in\mathbb{R}^d}$ with some symmetric positive definite $d\times d$-matrix $Q$ and fixed exponent $\alpha\in(1,2)$. The transition probability density of this process is the fundamental solution of some parabolic pseudo-differential equation. The notion of a single-layer potential for that equation is introduced and its properties are investigated. In particular, an operator is constructed whose role in our consideration is analogous to that the gradient in the classical theory. An analogy to the classical theorem on the jump of the co-normal derivative of the single-layer potential is proved. This result can be applied for solving some boundary-value problems for the parabolic pseudo-differential equations under consideration. For $\alpha=2$, the process under consideration is a linear transformation of Brownian motion, and all the investigated properties of the single-layer potential are well known.

Scale-invariant unconstrained online learning

We consider an online supervised learning problem, in which both the instances (input vectors) and the comparator (weight vector) are unconstrained. We exploit a natural scale invariance symmetry in our unconstrained setting: the predictions of the optimal comparator are invariant under any linear transformation of the instances. Our goal is to design online algorithms which also enjoy this property, i.e. are scale-invariant. We start with the case of coordinate-wise invariance, in which the individual coordinates (features) can be arbitrarily rescaled. We give an algorithm, which achieves essentially optimal regret bound in this setup, expressed by means of a coordinate-wise scale-invariant norm of the comparator. We then study general invariance with respect to arbitrary linear transformations. We first give a negative result, showing that no algorithm can achieve a meaningful bound in terms of scale-invariant norm of the comparator in the worst case. Next, we compliment this result with a positive one, providing an algorithm which “almost” achieves the desired bound, incurring only a logarithmic overhead in terms of the relative size of the instances.

8×8 reconfigurable quantum photonic processor based on silicon nitride waveguides.

The development of large-scale optical quantum information processing circuits ground on the stability and reconfigurability enabled by integrated photonics. We demonstrate a reconfigurable 8×8 integrated linear optical network based on silicon nitride waveguides for quantum information processing. Our processor implements a novel optical architecture enabling any arbitrary linear transformation and constitutes the largest programmable circuit reported so far on this platform. We validate a variety of photonic quantum information processing primitives, in the form of Hong-Ou-Mandel interference, bosonic coalescence/anti-coalescence and high-dimensional single-photon quantum gates. We achieve fidelities that clearly demonstrate the promising future for large-scale photonic quantum information processing using low-loss silicon nitride.

Complexity Analysis of Iterative Basis Transformations Applied to Event-Based Signals.

This paper introduces an event-based methodology to perform arbitrary linear basis transformations that encompass a broad range of practically important signal transforms, such as the discrete Fourier transform (DFT) and the discrete wavelet transform (DWT). We present a complexity analysis of the proposed method, and show that the amount of required multiply-and-accumulate operations is reduced in comparison to frame-based method in natural video sequences, when the required temporal resolution is high enough. Experimental results on natural video sequences acquired by the asynchronous time-based neuromorphic image sensor (ATIS) are provided to support the feasibility of the method, and to illustrate the gain in computation resources.

Designing optical circuits using plasmonic beam splitters.

Plasmonic optical circuits hold great promise in reducing device footprints by orders of magnitude and enabling high device complexity. The decomposition of an arbitrary linear transformation using unitary beam splitters is well-known. However, because of the inherent lossy nature of plasmonic devices, this decomposition is not useful for the practical design of devices. In this Letter, we provide a method to design an arbitrary unitary transformation using plasmonic beam splitters, which takes into account the inherent lossy nature of plasmonic modes in the decomposition process itself, while preserving the fidelity of the transformation. We do this by selecting the loss in each arm of the beam splitters and the interconnects. We also show how this method can be extended for the case of any linear transformation by extending the singular value decomposition. This method is applicable to plasmonic and waveguide-based lossy beam splitters.

Quantum Optical Realization of Arbitrary Linear Transformations Allowing for Loss and Gain

Unitary transformations are routinely modeled and implemented in the field of quantum optics. In contrast, nonunitary transformations that can involve loss and gain require a different approach. In this theory work, we present a universal method to deal with nonunitary networks. An input to the method is an arbitrary linear transformation matrix of optical modes that does not need to adhere to bosonic commutation relations. The method constructs a transformation that includes the network of interest and accounts for full quantum optical effects related to loss and gain. Furthermore, through a decomposition in terms of simple building blocks it provides a step-by-step implementation recipe, in a manner similar to the decomposition by Reck et al. [Reck et al., Phys. Rev. Lett. 73, 58 (1994)] but applicable to nonunitary transformations. Applications of the method include the implementation of positive-operator-valued measures and the design of probabilistic optical quantum information protocols.

Wigner functions on non-standard symplectic vector spaces

We consider the Weyl quantization on a flat non-standard symplectic vector space. We focus mainly on the properties of the Wigner functions defined therein. In particular we show that the sets of Wigner functions on distinct symplectic spaces are different but have non-empty intersections. This extends previous results to arbitrary dimension and arbitrary (constant) symplectic structure. As a by-product we introduce and prove several concepts and results on non-standard symplectic spaces which generalize those on the standard symplectic space, namely, the symplectic spectrum, Williamson’s theorem, and Narcowich-Wigner spectra. We also show how Wigner functions on non-standard symplectic spaces behave under the action of an arbitrary linear coordinate transformation.

Optimal Graph-Filter Design and Applications to Distributed Linear Network Operators

We study the optimal design of graph filters (GFs) to implement arbitrary linear transformations between graph signals. GFs can be represented by matrix polynomials of the graph-shift operator (GSO). Since this operator captures the local structure of the graph, GFs naturally give rise to distributed linear network operators. In most setups, the GSO is given so that GF design consists fundamentally in choosing the (filter) coefficients of the matrix polynomial to resemble desired linear transformations. We determine spectral conditions under which a specific linear transformation can be implemented perfectly using GFs. For the cases where perfect implementation is infeasible, we address the optimization of the filter coefficients to approximate the desired transformation. Additionally, for settings where the GSO itself can be modified, we study its optimal design as well. After this, we introduce the notion of a node-variant GF, which allows the simultaneous implementation of multiple (regular) GFs in different nodes of the graph. This additional flexibility enables the design of more general operators without undermining the locality in implementation. Perfect and approximate designs are also studied for this new type of GFs. To showcase the relevance of the results in the context of distributed linear network operators, this paper closes with the application of our framework to two particular distributed problems: finite-time consensus and analog network coding.

Implementing BosonSampling with time-bin encoding: Analysis of loss, mode mismatch, and time jitter

It was recently shown by Motes, Gilchrist, Dowling & Rohde [PRL 113, 120501 (2014)] that a time-bin encoded fiber-loop architecture can implement an arbitrary passive linear optics transformation. This was shown in the case of an ideal scheme whereby the architecture has no sources of error. In any realistic implementation, however, physical errors are present, which corrupt the output of the transformation. We investigate the dominant sources of error in this architecture --- loss and mode-mismatch --- and consider how it affects the BosonSampling protocol, a key application for passive linear optics. For our loss analysis we consider two major components that contribute to loss --- fiber and switches --- and calculate how this affects the success probability and fidelity of the device. Interestingly, we find that errors due to loss are not uniform (unique to time-bin encoding), which asymmetrically biases the implemented unitary. Thus, loss necessarily limits the class of unitaries that may be implemented, and therefore future implementations must prioritise minimising loss rates if arbitrary unitaries are to be implemented. Our formalism for mode-mismatch is generlized to account for various phenomenon that may cause mode-mismatch, but we focus on two --- errors in fiber-loop lengths, and time-jitter of the photon source. These results provide a guideline for how well future experimental implementations might perform in light of these error mechanisms.

Realization of arbitrary discrete unitary transformations using spatial and internal modes of light

Any lossless transformation on $n_{s}$ spatial and $n_{p}$ internal modes of light can be described by an $n_{s}n_{p}\times n_{s}n_{p}$ unitary matrix, but there is no known procedure to effect an arbitrary $n_{s}n_{p}\times n_{s}n_{p}$ unitary matrix on light in $n_{s}$ spatial and $n_{p}$ internal modes. We devise an algorithm to realize an arbitrary discrete unitary transformation on the combined spatial and internal degrees of freedom of light. Our realization uses beamsplitters and operations on internal modes to effect arbitrary linear transformations. The number of beamsplitters required to realize a unitary transformation is reduced as compared to existing realization by a factor $n_{p}^2/2$ at the cost of increasing the number of internal optical elements by a factor of two. Our algorithm thus enables the optical implementation of higher dimensional unitary transformations.

Redundancy of Moment Conditions for Linear Transformation of Parameters

In this paper, we consider the redundancy of an extra set of moment conditions, given an initial set of moment conditions, for the efficient estimation of an arbitrary linear transformation of an original parameter vector. The redundancy condition derived in the current paper unifies the full and partial redundancy of moment conditions of Breusch, Qian, Schmidt and Wyhowski (Journal of Econometrics, 1999).