Nanoscale neural network using non-linear spin-wave interference

We demonstrate the design of a neural network hardware, where all neuromorphic computing functions, including signal routing and nonlinear activation are performed by spin-wave propagation and interference. Weights and interconnections of the network are realized by a magnetic-field pattern that is applied on the spin-wave propagating substrate and scatters the spin waves. The interference of the scattered waves creates a mapping between the wave sources and detectors. Training the neural network is equivalent to finding the field pattern that realizes the desired input-output mapping. A custom-built micromagnetic solver, based on the Pytorch machine learning framework, is used to inverse-design the scatterer. We show that the behavior of spin waves transitions from linear to nonlinear interference at high intensities and that its computational power greatly increases in the nonlinear regime. We envision small-scale, compact and low-power neural networks that perform their entire function in the spin-wave domain.

Odd Willis coupling induced by broken time-reversal symmetry

When sound interacts with geometrically asymmetric structures, it experiences coupling between pressure and particle velocity, known as Willis coupling. While in most instances this phenomenon is perturbative in nature, tailored asymmetries combined with resonances can largely enhance it, enabling exotic acoustic phenomena. In these systems, Willis coupling obeys reciprocity, imposing an even symmetry of the Willis coefficients with respect to time reversal and the impinging wave vector, which translates into stringent constraints on the overall scattering response. In this work, we introduce and experimentally observe a dual form of acoustic Willis coupling, arising in geometrically symmetric structures when time-reversal symmetry is broken, for which the pressure-velocity coupling is purely odd-symmetric. We derive the conditions to maximize this effect, we experimentally verify it in a symmetric subwavelength scatterer biased by angular momentum, and we demonstrate the opportunities for sound scattering enabled by odd Willis coupling. Our study opens directions for acoustic metamaterials, with direct implications for sound control, non-reciprocal scattering, wavefront shaping and signal routing, of broad interest also for nano-optics, photonics, elasto-dynamics, and mechanics.

mmW Rotman Lens-Based Sensing: An Investigation Study

A Rotman lens is a wideband true-time delay device. Due to its simplistic structure with wave/signal routing capabilities, it has been widely utilized as a beamforming device in numerous communication systems. Since the basic Rotman lens design incorporates multiple input, output, and dummy ports, in this study, and for the first time, we utilized a Rotman lens as a sensor. The main idea was to gather abundant information from available Rotman lens ports to obtain better sensing performance. The realized lens is optimized to work in the millimeter wave (mmW) band from 27 to 29 GHz with a focus on a central frequency of 28 GHz. The design has a footprint of 140 × 103 × 0.8 mm3. The polarity correlator was used to characterize the material under investigation.

Multiplane Mesoscope reveals distinct cortical interactions following expectation violations

Cortical columns interact through dynamic routing of neuronal activity. To monitor these interactions, we developed the Multiplane Mesoscope which combines three established microscopy technologies: time-division multiplexing, remote focusing, and random-access mesoscopy. The Multiplane Mesoscope allowed us to study cortical column interactions in excitatory and inhibitory subpopulations in behaving mice. We found that distinct cortical subnetworks represent expected and unexpected events, suggesting that expectation violations modify signal routing across cortical columns, and establishing the Multiplane Mesoscope as a unique platform to study signal routing.

Measuring Directed Functional Connectivity Using Non-Parametric Directionality Analysis: Validation and Comparison with Non-Parametric Granger Causality

Background‘Non-parametric directionality’ (NPD) is a novel method for estimation of directed functional connectivity (dFC) in neural data. The method has previously been verified in its ability to recover causal interactions in simulated spiking networks in Halliday et al. (2015)MethodsThis work presents a validation of NPD in continuous neural recordings (e.g. local field potentials). Specifically, we use autoregressive model to simulate time delayed correlations between neural signals. We then test for the accurate recovery of networks in the face of several confounds typically encountered in empirical data. We examine the effects of NPD under varying: a) signal-to-noise ratios, b) asymmetries in signal strength, c) instantaneous mixing, d) common drive, e) and parallel/convergent signal routing. We also apply NPD to data from a patient who underwent simultaneous magnetoencephalography and deep brain recording.ResultsWe demonstrate that NPD can accurately recover directed functional connectivity from simulations with known patterns of connectivity. The performance of the NPD metric is compared with non-parametric Granger causality (NPG), a well-established methodology for model free estimation of dFC. A series of simulations investigating synthetically imposed confounds demonstrate that NPD provides estimates of connectivity that are equivalent to NPG. However, we provide evidence that: i) NPD is less sensitive than NPG to degradation by noise; ii) NPD is more robust to the generation of false positive identification of connectivity resulting from SNR asymmetries; iii) NPD is more robust to corruption via moderate degrees of instantaneous signal mixing.ConclusionsThe results in this paper highlight that to be practically applied to neural data, connectivity metrics should not only be accurate in their recovery of causal networks but also resistant to the confounding effects often encountered in experimental recordings of multimodal data. Taken together, these findings position NPD at the state-of-the-art with respect to the estimation of directed functional connectivity in neuroimaging.HighlightsNon-parametric directionality (NPD) is a novel directed connectivity metric.NPD estimates are equivalent to Granger causality but more robust to signal confounds.Multivariate extensions of NPD can correctly identify signal routing.AbbreviationsdFCDirected functional connectivityEEGElectroencephalogramLFPLocal field potentialMEGMagnetoencephalogramMVARMultivariate autoregressive (model)NPDNon-parametric directionalityNPGNon-parametric Granger (causality)SMASupplementary motor areaSNRSignal-to-noise ratioSTNSubthalamic Nucleus

Causally investigating cortical dynamics and signal processing by targeting natural system attractors with precisely timed stimulation

1AbstractElectrical stimulation is a promising tool for interacting with neuronal dynamics to identify neural mechanisms that underlie cognitive function. Since effects of a single short stimulation pulse typically vary greatly and depend on the current network state, many experimental paradigms have rather resorted to continuous or periodic stimulation in order to establish and maintain a desired effect. However, such an approach explicitly leads to forced and ‘unnatural’ brain activity. Further, continuous stimulation can make it hard to parse the recorded activity and separate neural signal from stimulation artifacts. In this study we propose an alternate strategy: by monitoring a system in realtime, we use the existing preferred states or attractors of the network and to apply short and precise pulses in order to switch between its preferred states. When pushed into one of its attractors, one can use the natural tendency of the system to remain in such a state to prolong the effect of a stimulation pulse, opening a larger window of opportunity to observe the consequences on cognitive processing. To elaborate on this idea, we consider flexible information routing in the visual cortex as a prototypical example. When processing a stimulus, neural populations in the visual cortex have been found to engage in synchronized gamma activity. In this context, selective signal routing is achieved by changing the relative phase between oscillatory activity in sending and receiving populations (communication through coherence, CTC). In order to explore how perturbations interact with CTC, we investigate a biophysically realistic network exhibiting similar synchronization and signal routing phenomena. We develop a closed-loop stimulation paradigm based on the phase-response characteristics of the network and demonstrate its ability to establish desired synchronization states. By measuring information content throughout the model, we evaluate the effect of signal contamination caused by the stimulation in relation to the magnitude of the injected pulses and intrinsic noise in the system. Finally, we demonstrate that, up to a critical noise level, precisely timed perturbations can be used to artificially induce the effect of attention by selectively routing visual signals to higher cortical areas.