arXiv Science⌕ Search

arXiv subjects

George Sarantoglou

Publications and source records attributed to George Sarantoglou.

10 recordsLinked to original sources
arXiv2025

Reconfigurable Silicon Photonics Extreme Learning Machine with Random Non-linearities as Neural Processor and Physical Unclonable Function

An alternative extreme learning machine -ELM- paradigm is presented exploiting random non-linearities -RN, named RN-ELM, instead of a conventional fixed node non-linearity. This method is implemented on a hybrid neural engine, with the physical layer realized by an integrated silicon photonic mesh and the digital layer by a simple regression algorithm. Non-linearities are intrinsically non-power depended and are generated through non-linear frequency to power mapping offered by optical filters. The numerical evaluation is based on an experimentally derived transfer function of an all-pass filter, implemented on a silicon reconfigurable photonic integrated chip -RPIC. RN-ELM is evaluated in a twofold manner; first as a machine learning scheme, where the expressivity offered by multiple, yet random, activation functions lead to a compact and highly simplified design with 5 optical filters, offering state-of-the-art performance in time-series prediction tasks with minimum hardware requirements. The second scenario entails its deployment as a physical unclonable function -PUF, for authentication applications directly in the physical layer. In this case, the random activation functions are associated with unavoidable, fabrication related waveguide imperfections that can act as hardware signatures. Numerical results reveal a probability of cloning as low as 10e-15, which corresponds to a highly secure authentication token.

physics.optics↗
arXiv2025

Reconfigurable Integrated Photonic Chips as Dual-Purpose Neuromorphic Accelerators and Physical Unclonable Functions

In this work, we experimentally validate the dual use of a reconfigurable photonic integrated mesh as a neuromorphic accelerator, targeting signal equalization, and as a physical unclonable function offering authentication at the hardware level. The processing node is an optical spectrum slicing self-coherent transceiver targeting the mitigation of dispersion impairments of an intensity detected QPSK signal, after 25 km of transmission at 32 Gbaud. Unavoidable fabrication related imperfections of the nodes, such as waveguide roughness, can act as fingerprints of the device, and, during neuromorphic processing, result in unique weights at the digital back-end during signal equalization. Extracted security metrics offer low false positive/negative probability for the generated responses, confirming un-clonability, whereas bit-error-ratio for the QPSK equalization task was always below the hardware forward error correction limit. The experimental results substantiate the capability of the proposed scheme to simultaneously act as an accelerator and as a security token.

physics.optics↗
arXiv2025

Experimental Analysis of a Self-Coherent M-QAM Receiver by Means of Recurrent Optical Spectrum Slicing and Direct Detection

High order modulation formats constitute the most prominent way for increasing spectral efficiency in transmission systems. Coherent transceivers that support such higher order formats require heavy digital signal processing (DSP), which increases the power consumption of coherent pluggables, well above the intensity modulation and direct detection (IM/DD) counterparts. Self-coherent or phase retrieval methods have emerged as potential solutions, trying to combine the merits of coherent technology with the simplicity of direct detection. In this work, we experimentally demonstrate the reception of quadrature amplitude modulation (QAM) modulation formats based on direct detection aided by the recurrent optical spectrum slicing (ROSS) photonic accelerator, utilizing minimal DSP and low modulator driving voltages. We experimentally demonstrate 32 Gbaud QAM-4/16 for 25 km, 50 km and 75 km in the C-band aided by a linear digital equalization and the use of programmable photonics as recurrent optical spectrum slicers. We showcase successful detection with driving swings below V{\pi}/3 in contrast to the full swing required by conventional coherent transceivers. We further improve the system performance utilizing geometric constellation shaping. Finally, we explore the potential power consumption improvement for the next-generation 1.6T pluggables, showcasing over 40% reduction with respect to the most lightweight state of the art coherent solutions reported in literature

physics.optics↗
arXiv2024

Photonic Neuromorphic Accelerators for Event-Based Imaging Flow Cytometry

In this work, we present experimental results of a high-speed label-free imaging cytometry system that seamlessly merges the high-capturing rate and data sparsity of an event-based CMOS camera with lightweight photonic neuromorphic processing. This combination offers high classification accuracy and a massive reduction in the number of trainable parameters of the digital machine-learning back-end. The photonic neuromorphic accelerator is based on a hardware-friendly passive optical spectrum slicing technique that is able to extract meaningful features from the generated spike-trains. The experimental scenario comprises the discrimination of artificial polymethyl methacrylate calibrated beads, having different diameters, flowing at a mean speed of 0.01m/sec. Classification accuracy, using only lightweight, digital machine-learning schemes has topped at 98.2%. On the other hand, by experimentally pre-processing the raw spike data through the proposed photonic neuromorphic spectrum slicer we achieved an accuracy of 98.6%. This performance was accompanied by a reduction in the number of trainable parameters at the classification back-end by a factor ranging from 8 to 22, depending on the configuration of the digital neural network.

physics.optics↗
arXiv2024

Experimental Investigation of a Recurrent Optical Spectrum Slicing Receiver for Intensity Modulation/Direct Detection systems using Programmable Photonics

In this paper, we experimentally validate our previous numerical works in recurrent optical spectrum slicing (ROSS) accelerators for dispersion compensation in high-speed IM/DD links. For this, we utilize recurrent filters implemented both through a waveshaper and by exploiting novel integrated programmable photonic platforms. Different recurrent configurations are tested. The ROSS accelerators exploit frequency processing through recurrent optical filter nodes in order to mitigate the power fading effect, which hinders the transmission distance and baudrate scalability of IM/DD systems. By equalizing even 80 km of 64 Gb/s PAM-4 transmission in C-band, we prove that our system can offer an appealing solution in highly dispersive channels.

physics.optics↗
arXiv2024

All-Optical, Reconfigurable and Power Independent Neural Activation Function by Means of Phase Modulation

In this work, we present numerical results concerning an integrated photonic non-linear activation function that relies on a power independent, non-linear phase to amplitude conversion in a passive optical resonator. The underlying mechanism is universal to all optical filters, whereas here, simulations were based on micro-ring resonators (MRRs). Investigation revealed that the photonic neural node can be tuned to support a wide variety of continuous activation functions that are relevant to the neural network architectures, such as the sigmoid and the softplus functions. The proposed photonic node is numerically evaluated in the context of time delayed reservoir computing (TDRC) scheme, targeting the one-step ahead prediction of the Santa Fe series. The proposed phase to amplitude TDRC is benchmarked versus the conventional amplitude based TDRC, showcasing a performance boost by one order of magnitude.

physics.optics↗
arXiv2023

Integrated Photonic Accelerator Based on Optical Spectrum Slicing for Convolutional Neural Networks

In this work we numerically analyze a passive photonic integrated neuromorphic accelerator based on hardware-friendly optical spectrum slicing nodes. The proposed scheme can act as a fully analogue convolutional layer, preprocessing information directly in the optical domain. The proposed scheme allows the extraction of meaningful spatio-temporal features from the incoming data, thus when used prior to a simple fully connected digital single layer network it can boost performance with negligible power consumption. Numerical simulations using the MNIST dataset confirmed the acceleration properties of the proposed scheme, where 10 neuromorphic nodes can replace the convolutional layers of a sophisticated LeNet-5 network, thus reducing the number of total floating point operations per second (FLOPS) by 98% while offering a 97.2% classification accuracy.

cs.ET↗
arXiv2022

Bayesian Photonic Accelerators for Energy Efficient and Noise Robust Neural Processing

Artificial neural networks are efficient computing platforms inspired by the brain. Such platforms can tackle a vast area of real-life tasks ranging from image processing to language translation. Silicon photonic integrated chips (PICs), by employing coherent interactions in Mach-Zehnder interferometers, are promising accelerators offering record low power consumption and ultra-fast matrix multiplication. Such photonic accelerators, however, suffer from phase uncertainty due to fabrication errors and crosstalk effects that inhibit the development of high-density implementations. In this work, we present a Bayesian learning framework for such photonic accelerators. In addition to the conventional log-likelihood optimization path, two novel training schemes are derived, namely a regularized version and a fully Bayesian learning scheme. They are applied on a photonic neural network with 512 phase shifters targeting the MNIST dataset. The new schemes, when combined with a pre-characterization stage that provides the passive offsets, are able to dramatically decrease the operational power of the PIC beyond 70%, with just a slight loss in classification accuracy. The full Bayesian scheme, apart from this energy reduction, returns information with respect to the sensitivity of the phase shifters. This information is used to de-activate 31% of the phase actuators and, thus, significantly simplify the driving system.

cs.ET↗