Neuromorphic computation with a single magnetic domain wall.

Machine learning techniques are commonly used to model complex relationships but implementations on digital hardware are relatively inefficient due to poor matching between conventional computer architectures and the structures of the algorithms they are required to simulate. Neuromorphic devices, and in particular reservoir computing architectures, utilize the inherent properties of physical systems to implement machine learning algorithms and so have the potential to be much more efficient. In this work, we demonstrate that the dynamics of individual domain walls in magnetic nanowires are suitable for implementing the reservoir computing paradigm in hardware. We modelled the dynamics of a domain wall placed between two anti-notches in a nickel nanowire using both a 1D collective coordinates model and micromagnetic simulations. When driven by an oscillating magnetic field, the domain exhibits non-linear dynamics within the potential well created by the anti-notches that are analogous to those of the Duffing oscillator. We exploit the domain wall dynamics for reservoir computing by modulating the amplitude of the applied magnetic field to inject time-multiplexed input signals into the reservoir, and show how this allows us to perform machine learning tasks including: the classification of (1) sine and square waves; (2) spoken digits; and (3) non-temporal 2D toy data and hand written digits. Our work lays the foundation for the creation of nanoscale neuromorphic devices in which individual magnetic domain walls are used to perform complex data analysis tasks.

Information-theoretic analysis of iterated Bayesian acoustic source localization in a static ocean waveguide.

Fundamental constructs of information theory are applied to quantify the performance of iterated (sequential) Bayesian localization of a time-harmonic source in a range- and time-invariant acoustic waveguide using the segmented Fourier transforms of the received pressure time series. The nonlinear relation, defined by acoustic propagation, between the source location and the received narrowband spectral components is treated as a nonlinear communication channel. The performance analysis includes mismatch between the acoustic channel and the model channel on which the Bayesian inference is based. Source location uncertainty is quantified by the posterior probability density of source location, by the posterior entropy and associated uncertainty area, by the information gain (relative entropy) at each iteration, and by large-ensemble limits of these quantities. A computational example for a vertical receiver array in a shallow-water waveguide is presented with acoustic propagation represented by normal modes and ambient noise represented by a Kuperman-Ingenito model. Performance degradation due to noise-model mismatch is quantified in an example. Potential extensions to uncertain and stochastic environments are discussed.

Empirical and quadrature approximation of acoustic field and array response probability density functions.

Numerical methods are presented for approximating the probability density functions (pdf's) of acoustic fields and receiver-array responses induced by a given joint pdf of a set of acoustic environmental parameters. An approximation to the characteristic function of the random acoustic field (the inverse Fourier transform of the field pdf) is first obtained either by construction of the empirical characteristic function (ECF) from a random sample of the acoustic parameters, or by application of generalized Gaussian quadrature to approximate the integral defining the characteristic function. The Fourier transform is then applied to obtain an approximation of the pdf by a continuous function of the field variables. Application of both the ECF and generalized Gaussian quadrature is demonstrated in an example of a shallow-water ocean waveguide with two-dimensional uncertainty of sound speed and attenuation coefficient in the ocean bottom. Both approximations lead to a smoother estimate of the field pdf than that provided by a histogram, with generalized Gaussian quadrature providing a smoother estimate at the tails of the pdf. Potential applications to acoustic system performance quantification and to nonparametric acoustic signal processing are discussed.

Realization of the manipulation of ultracold atoms with a reconfigurable nanomagnetic system of domain walls.

Planar magnetic nanowires have been vital to the development of spintronic technology. They provide an unparalleled combination of magnetic reconfigurability, controllability, and scalability, which has helped to realize such applications as racetrack memory and novel logic gates. Microfabricated atom optics benefit from all of these properties, and we present the first demonstration of the amalgamation of spintronic technology with ultracold atoms. A magnetic interaction is exhibited through the reflection of a cloud of (87)Rb atoms at a temperature of 10 µK, from a 2 mm × 2 mm array of nanomagnetic domain walls. In turn, the incident atoms approach the array at heights of the order of 100 nm and are thus used to probe magnetic fields at this distance.

Acoustic field and array response uncertainties in stratified ocean media.

The change-of-variables theorem of probability theory is applied to compute acoustic field and array beam power probability density functions (pdfs) in uncertain ocean environments represented by stratified, attenuating ocean waveguide models. Computational studies for one and two-layer waveguides investigate the functional properties of the acoustic field and array beam power pdfs. For the studies, the acoustic parameter uncertainties are represented by parametric pdfs. The field and beam response pdfs are computed directly from the parameter pdfs using the normal-mode representation and the change-of-variables theorem. For two-dimensional acoustic parameter uncertainties of sound speed and attenuation, the field and beam power pdfs exhibit irregular functional behavior and singularities associated with stationary points of the mapping, defined by acoustic propagation, from the parameter space to the field or beam power space. Implications for the assessment of orthogonal polynomial expansion and other methods for computing acoustic field pdfs are discussed.

Single- and multi-channel underwater acoustic communication channel capacity: a computational study.

Acoustic communication channel capacity determines the maximum data rate that can be supported by an acoustic channel for a given source power and source/receiver configuration. In this paper, broadband acoustic propagation modeling is applied to estimate the channel capacity for a time-invariant shallow-water waveguide for a single source-receiver pair and for vertical source and receiver arrays. Without bandwidth constraints, estimated single-input, single-output (SISO) capacities approach 10 megabitss at 1 km range, but beyond 2 km range they decay at a rate consistent with previous estimates by Peloquin and Leinhos (unpublished, 1997), which were based on a sonar equation calculation. Channel capacities subject to source bandwidth constraints are approximately 30-90% lower than for the unconstrained case, and exhibit a significant wind speed dependence. Channel capacity is investigated for single-input, multi-output (SIMO) and multi-input, multi-output (MIMO) systems, both for finite arrays and in the limit of a dense array spanning the entire water column. The limiting values of the SIMO and MIMO channel capacities for the modeled environment are found to be about four times higher and up to 200-400 times higher, respectively, than for the SISO case. Implications for underwater acoustic communication systems are discussed.