arXiv subjects

Ilya Starshynov

Publications and source records attributed to Ilya Starshynov.

14 recordsLinked to original sources

arXiv2025

Fisher information flow in artificial neural networks

The estimation of continuous parameters from measured data plays a central role in many fields of physics. A key tool in understanding and improving such estimation processes is the concept of Fisher information, which quantifies how information about unknown parameters propagates through a physical system and determines the ultimate limits of precision. With Artificial Neural Networks (ANNs) gradually becoming an integral part of many measurement systems, it is essential to understand how they process and transmit parameter-relevant information internally. Here, we present a method to monitor the flow of Fisher information through an ANN performing a parameter estimation task, tracking it from the input to the output layer. We show that optimal estimation performance corresponds to the maximal transmission of Fisher information, and that training beyond this point results in information loss due to overfitting. This provides a model-free stopping criterion for network training-eliminating the need for a separate validation dataset. To demonstrate the practical relevance of our approach, we apply it to a network trained on data from an imaging experiment, highlighting its effectiveness in a realistic physical setting.

cs.LG↗

arXiv2025

Model-free estimation of the Cramér-Rao bound for deep-learning microscopy in complex media

Artificial neural networks have become important tools to harness the complexity of disordered or random photonic systems. Recent applications include the recovery of information from light that has been scrambled during propagation through a complex scattering medium, especially in the challenging case where the deterministic input-output transmission matrix cannot be measured. This naturally raises the question of what the limit is that information theory imposes on this recovery process, and whether neural networks can actually reach this limit. To answer these questions, we introduce a model-free approach to calculate the Cramér-Rao bound, which sets the ultimate precision limit at which artificial neural networks can operate. As an example, we apply this approach in a proof-of-principle experiment using laser light propagating through a disordered medium, evidencing that a convolutional network approaches the ultimate precision limit in the challenging task of localizing a reflective target hidden behind a dynamically-fluctuating scattering medium. The model-free method introduced here is generally applicable to benchmark the performance of any deep-learning microscope, to drive algorithmic developments and to push the precision of metrology and imaging techniques to their ultimate limit.

physics.optics↗

arXiv2024

Photon transport through the entire adult human head

Optical brain imaging technologies are promising due to their relatively high temporal resolution, portability and cost-effectiveness. However, the highly scattering nature of near-infrared light in human tissue makes it challenging to collect photons emerging from more than 4 cm below the scalp, or with source-detector separation larger than several centimeters. We explore the physical limits of photon transport in the head and show that despite an extreme attenuation of ~10^(18), we can experimentally detect light that is transmitted diametrically through the entire adult human head. Analysis of various photon migration pathways through the head also indicates how the source-detector configuration can be used to isolate photons interacting with deep regions of the brain that are inaccessible with current optical techniques.

physics.optics↗

arXiv2024

AI-Enabled sensor fusion of time of flight imaging and mmwave for concealed metal detection

In the field of detection and ranging, multiple complementary sensing modalities may be used to enrich the information obtained from a dynamic scene. One application of this sensor fusion is in public security and surveillance, whose efficacy and privacy protection measures must be continually evaluated. We present a novel deployment of sensor fusion for the discrete detection of concealed metal objects on persons whilst preserving their privacy. This is achieved by coupling off-the-shelf mmWave radar and depth camera technology with a novel neural network architecture that processes the radar signals using convolutional Long Short-term Memory (LSTM) blocks and the depth signal, using convolutional operations. The combined latent features are then magnified using a deep feature magnification to learn cross-modality dependencies in the data. We further propose a decoder, based on the feature extraction and embedding block, to learn an efficient upsampling of the latent space to learn the location of the concealed object in the spatial domain through radar feature guidance. We demonstrate the detection of presence and inference of 3D location of concealed metal objects with an accuracy of up to 95%, using a technique that is robust to multiple persons. This work provides a demonstration of the potential for cost effective and portable sensor fusion, with strong opportunities for further development.

eess.IV↗

arXiv2024

Energy transport in diffusive waveguides

The guiding and transport of energy, for example of electromagnetic waves underpins many technologies that have shaped modern society, ranging from long distance optical fibre telecommunications to on-chip optical processors. Traditionally, a mechanism is required that exponentially localises the waves or particles in the confinement region, e.g. total internal reflection at a boundary. We introduce a waveguiding mechanism that relies on a different origin for the exponential confinement and that arises due to the physics of diffusion. We demonstrate this concept using light and show that photon density waves can propagate as a guided mode along a core-structure embedded in a scattering, opaque material, enhancing light transmission by orders of magnitude and along non-trivial, e.g. curved trajectories. This waveguiding mechanism can also occur naturally, for example in the cerebral spinal fluid surrounding the brain, along tendons in the human body and is to be expected in other systems that follow the same physics e.g. neutron diffusion.

physics.optics↗

arXiv2022

Remote laser-speckle sensing of heart sounds for health assessment and biometric identification

Assessment of heart sounds is the cornerstone of cardiac examination, but it requires a stethoscope, skills and experience, and a direct contact with the patient. We developed a contactless, machine-learning assisted method for heart-sound identification and quantification based on the remote measurement of the reflected laser speckle from the neck skin surface in healthy individuals. We compare the performance of this method to standard digital stethoscope recordings on an example task of heart-beat sound biometric identification. We show that our method outperforms the stethoscope even allowing identification on the test data taken on different days. This method might allow development of devices for remote monitoring of cardiovascular health in different settings.

physics.med-ph↗

arXiv2022

Statistical Dependencies Beyond Linear Correlations in Light Scattered by Disordered Media

Imaging through scattering and random media is an outstanding problem that to date has been tackled by either measuring the medium transmission matrix or exploiting linear correlations in the transmitted speckle patterns. However, transmission matrix techniques require interferometric stability and linear correlations, such as the memory effect, can be exploited only in thin scattering media. Here we show the existence of a statistical dependency in strongly scattered optical fields in a case where first-order correlations are not expected. We also show that this statistical dependence and the related information transport is directly linked to artificial neural network imaging in strongly scattering, dynamic media. These non-trivial dependencies provide a key to imaging through dynamic and thick scattering media with applications for deep-tissue imaging or imaging through smoke or fog

physics.optics↗

arXiv2021

Measurement of Penrose superradiance in a photon superfluid

The superradiant amplification in the scattering from a rotating medium was first elucidated by Sir Roger Penrose over 50 years ago as a means by which particles could gain energy from rotating black holes. Despite this fundamental process being ubiquitous also in wave physics, it has only been observed once experimentally, in a water tank, and never in an astrophysical setting. Here, we measure this amplification for a nonlinear optics experiment in the superfluid regime. In particular, by focusing a weak optical beam carrying orbital angular momentum onto the core of a strong pump vortex beam, negative norm modes are generated and trapped inside the vortex core, allowing for amplification of a reflected beam. Our experiment demonstrates amplified reflection due to a novel form of nonlinear optical four-wave mixing, whose phase-relation coincides with the Zel'dovich-Misner condition for Penrose superradiance in our photon superfluid, and unveil the role played by negative frequency modes in the process.}

physics.optics↗

arXiv2021

3D imaging from multipath temporal echoes

Echo-location is a broad approach to imaging and sensing that includes both man-made RADAR, LIDAR, SONAR and also animal navigation. However, full 3D information based on echo-location requires some form of scanning of the scene in order to provide the spatial location of the echo origin-points. Without this spatial information, imaging objects in 3D is a very challenging task as the inverse retrieval problem is strongly ill-posed. Here, we show that the temporal information encoded in the return echoes that are reflected multiple times within a scene is sufficient to faithfully render an image in 3D. Numerical modelling and an information theoretic perspective prove the concept and provide insight into the role of the multipath information. We experimentally demonstrate the concept by using both radio-frequency and acoustic waves for imaging individuals moving in a closed environment.

eess.IV↗

arXiv2020

Imaging entanglement correlations with a single-photon avalanche diode camera

Spatial correlations between two photons are the key resource in realising many quantum imaging schemes. Measurement of the bi-photon correlation map is typically performed using single-point scanning detectors or single-photon cameras based on CCD technology. However, both approaches are limited in speed due to the slow scanning and the low frame-rate of CCD-based cameras, resulting in data acquisition times on the order of many hours. Here we employ a high frame rate, single photon avalanche diode (SPAD) camera, to measure the spatial joint probability distribution of a bi-photon state produced by spontaneous parametric down-conversion, with statistics taken over $10^7$ frames acquired in just 140 seconds. We verified the presence of spatial entanglement between our photon pairs through the violation of an Einstein-Podolsky-Rosen criterion, with a confidence level of 227 sigmas. Our work demonstrates the potential of SPAD cameras in the rapid characterisation of photon correlations, leading the way towards quantum imaging in real-time.

quant-ph↗

arXiv2020

Spatial images from temporal data

Traditional paradigms for imaging rely on the use of a spatial structure, either in the detector (pixels arrays) or in the illumination (patterned light). Removal of the spatial structure in the detector or illumination, i.e., imaging with just a single-point sensor, would require solving a very strongly ill-posed inverse retrieval problem that to date has not been solved. Here, we demonstrate a data-driven approach in which full 3D information is obtained with just a single-point, single-photon avalanche diode that records the arrival time of photons reflected from a scene that is illuminated with short pulses of light. Imaging with single-point time-of-flight (temporal) data opens new routes in terms of speed, size, and functionality. As an example, we show how the training based on an optical time-of-flight camera enables a compact radio-frequency impulse radio detection and ranging transceiver to provide 3D images.

eess.IV↗

arXiv2019

Coherent control of light for non-line-of-sight imaging

Non-line-of-sight (NLOS) imaging relies on collecting light that is rendered incoherent from the multiple scattering events and is then post-processed to provide an estimate of the hidden scene. Here we employ coherent phase control of the outgoing laser beam phase front so as to refocus the beam behind the obscuring obstacle and then use the speckle memory effect to scan the focused spot across the scene. The back-reflected light intensity provides a direct measurement of the scene with a signal-to-noise ratio that is greatly improved when measured using a temporally gated detector. A spatial resolution of less than 1 mm is demonstrated, opening the way to high-resolution NLOS imaging.

physics.optics↗

arXiv2018

Blind Ghost Imaging

Ghost imaging is an unconventional optical imaging technique that reconstructs the shape of an object combining the measurement of two signals: one that interacted with the object, but without any spatial information, the other containing spatial information, but that never interacted with the object. Ghost imaging is a very flexible technique, that has been generalized to the single-photon regime, to the time domain, to infrared and terahertz frequencies, and many more conditions. Here we demonstrate that ghost imaging can be performed without ever knowing the patterns illuminating the object, but using patterns correlated with them, doesn't matter how weakly. As an experimental proof we exploit the recently discovered correlation between the reflected and transmitted light from a scattering layer, and reconstruct the image of an object hidden behind a scattering layer using only the reflected light, which never interacts with the object. This method opens new perspectives for non-invasive imaging behind or within turbid media.

physics.optics↗

arXiv2015

Quantum correlation of light scattered by disordered media

We study theoretically how multiple scattering of light in a disordered medium can spontaneously generate quantum correlations. In particular we focus on the case where the input state is Gaussian and characterize the correlations between two arbitrary output modes. As there is not a single all-inclusive measure of correlation, we characterise the output correlations with three measures: intensity fluctuations, entanglement, and quantum discord. We found that, while a single mode coherent state input can not produce quantum correlations, any other Gaussian input will produce them in one form or another. This includes input states that are usually regarded as more classical than coherent ones, such as thermal states, which will produce a non zero quantum discord.

quant-ph↗