The prowess of Artificial Intelligence (AI) and Large Language Models (LLMs) to compute the exponentially growing data, coupled with the digital electronic AI computing systems reaching their physical plateaus, have stimulated extensive research into next-gen AI accelerators. Optical Neural Networks (ONNs) have been among the most favored candidates, since the unparalleled speed of light can provide massive compute parallelism by the physical dimensions of time, wavelength, and space, offering higher computational speeds and lower power consumption. In this paper, we experimentally demonstrate novel time-space-wavelength multiplexed arrayed waveguided gratings router (AWGR)-based architectures exploited as photonic matrix- and tensor-multiplier engines achieving a total computational power of up to 163.8 TOPS. This marks a substantial ∼15x improvement over state-of-the-art waveguide-based optical accelerators. The two demonstrators were tested for three different NN applications, under two experimental testbeds. The matrix-multiplier was evaluated as a fully connected NN (FCNN), classifying images from the FMNIST dataset. An 8×8 AWGR was employed for providing the passive routing, with modulators driven at 10Gbaud, exhibiting a hardware-based inference accuracy of 87.1% with respect to the software. The tensor-multiplier, using a 16x16 AWGR, was evaluated as a FCNN and a convolutional NN (CNN), detecting DDoS attacks and classifying handwritten digits at 20 Gbaud. The experimental results exhibited a Cohen's kappa score of 0.799 for the DDoS detection and a classification accuracy of 93.35%, respectively.

The ever-increasing volume of data demarcating from the exponential scale of Artificial Intelligence (AI) and Deep Learning (DL) models motivated research into specialized AI accelerators in order to complement digital processors. Photonic Neural Networks (PNNs), with their unique ability to capitalize on the interplay of multiple physical dimensions, including time, wavelength, and space, have been brought forward with a credible promise for boosting computational power and energy efficiency in AI processors. In this article, we experimentally demonstrate a novel multidimensional arrayed waveguide grating router (AWGR)-based photonic AI accelerator that can offload bandwidth-bounded linear algebra while leaving memory hierarchy, control, and nonlinearities to electronics and can execute tensor multiplications at a record-high total computational power of 262 TOPS, offering a ∼24× improvement over the existing waveguide-based optical accelerators. It consists of a 16 × 16 AWGR that exploits the time-, wavelength-, and space-division multiplexing (T-W-SDM) for weight and input encoding, together with an integrated Si3N4-based frequency comb for multi-wavelength generation. The photonic AI accelerator has been experimentally validated in both Fully Connected (FC) and Convolutional NN (CNN) models, with the FC and CNN being trained for DDoS attack identification and MNIST classification, respectively. The experimental inference at 32 Gbaud achieved a Cohen’s kappa score of 0.8652 for DDoS detection and an accuracy of 92.14% for MNIST classification, respectively, closely matching the software performance.

Photonic neural networks (PNNs) are projected to be the next-generation AI platform, driving significant advances in compute energy and area efficiency by leveraging light’s parallelism across space, wavelength, and time. Here we present a photonic integrated tensor core (PITC) architecture that combines an arrayed waveguide grating router (AWGR)-enabled PNN with a SiGe electro-absorption modulator (EAM) array following a chiplet-based silicon photonic (SiPho) accelerator setup for multi-wavelength data and weight encoding. The 8 × 8 PITC architectural layout was experimentally validated in two AI benchmark workloads at 20 Gbaud, successfully classifying the IRIS and MNIST datasets yielding Cohen’s kappa-scores of 0.8438 and 0.7421, respectively.

The threads of photonics are eagerly awaited to redefine the future of neuromorphic data processing, especially as computing-intensive artificial intelligence models become an unavoidable part of our everyday lives. However, there is much to be improved within the domain of photonic nonlinear activation functions, as programmable, all-optical, energy-efficient nonlinearities remain beyond the grasp of the current state-of-the-art. In this paper, we address the issue at hand and propose a novel approach in the realization of high-performance all-optical photonic activations. Through simulations and experiments, we show that Fabry–Pérot laser diodes (FP-LDs) exhibit richness and high programmability of their nonlinear response to input optical pulses with widths as low as 25 ps. We demonstrate a variety of sigmoid-like and inverted PReLU-like trends to be used as all-optical activation functions in photonic neural networks, testing their performance in stringent, real-life training scenarios with randomized data patterns at repetition rates of up to 10 GHz. The programmability of activations is shown using a multitude of experimental operating parameters, among which we highlight the power variation of an additional continuous-wave laser injected into the FP-LD, enriching our approach with all-optical control of all-optical activations. With very low static power consumption of our active element, we achieve a record-breaking energy draw on the order of pJ to hundreds of fJ per nonlinear operation.

The never-ending evolution of the Internet of Things ecosystem is reshaping the arena of wireless communications and competing against conventional networking solutions in fields such as battery life, device and deployment cost, coverage, and support for an immense number of devices. Inspired by this phenomenon, this paper presents a novel Medium Access Control protocol utilizing long-range technology, based on a Time Division Multiple Access communication protocol variant, adjusted to make better use of each device’s hardware. Focusing on Low Power Wide Area Network applications, this implementation improves data latency and offers amplified performance due to better network awareness and dynamic time slot rescheduling. Various simulation scenarios were contrived to evaluate the protocol’s performance. The results instate the proposed algorithm as a promising access scheme for the IoT field.