Hardware-efficient photonic tensor core: accelerating deep neural networks with structured compression

In recent years, with the expansion of information, artificial intelligence technology has been developed and used in various fields. Among them, optical neural network provides a new type of special neural network accelerator chip solution, which has the advantages of high speed, high bandwidth, and low power consumption. In this paper, we construct an optical neural network based on Mach–Zehnder interferometer. The experimental results on the image classification of MNIST handwritten digitals show that the optical neural network has high accuracy, fast convergence and good scalability.

Power efficient ultra-wideband based wireless communication for wireless body area network applications

The rise of artificial intelligence (AI) has driven increasingly stringent requirements for computing speed and energy efficiency. Optical computing offers a promising alternative due to its inherent parallelism and low power consumption. Optical neural networks (ONNs), as an actively studied application of optical computing in AI, have shown potential in matrix operations and neural network inference. However, existing ONNs remain constrained by limited diversity of network layers and an insufficient number of tunable weights. Here, we propose the PCM-MD (phase-change-materials-microring-diffraction) hybrid ONN architecture based on microring resonators, diffraction slots, and phase-change materials that enables diverse optical parameterized layers. The architecture features 679 tunable weights distributed across convolutional, nonlinear, compression, and fully connected layers. PCM-MD demonstrates its effectiveness across multiple intelligent tasks, achieving 85.12% accuracy on Fashion-MNIST classification, 100% accuracy on 3-bit arithmetic addition, and enabling content generation by serving as both generator and discriminator in optical generative adversarial networks (GANs). This optical parameterized architecture provides a promising foundation for energy-efficient AI systems.

A novel configuration of an optical lateral inhibitory neural network using a two-dimensional internal-light-coupled optical device (ILCOD) array is proposed. It is experimentally certified with a test ILCOD structured with seven units in a two-dimensional hexagonal arrangement that the optical coupling between the units causes turn-off of the light emission of the unit in a specific position, which results in lateral inhibitory action of a neural network.

The GPU has provide higher throughput by integrating more execution resources into a single chip without unduly compromising power efficiency. With the power wall challenge, however, increasing the throughput will require significant improvement in power efficiency. To accomplish this goal, we propose G-Scalar, a cost-effective generalized scalar execution architecture for GPUs in this paper. G-Scalar offers two key advantages over prior architectures supporting scalar execution for only non-divergent arithmetic/logic instructions. First, G-Scalar is more power-efficient as it can also support scalar execution of divergent and special-function instructions, the fraction of which in contemporary GPU applications has notably increased. Second, G-Scalar is less expensive as it can share most of its hardware resources with register value compression, of which adoption has been strongly promoted to reduce high power consumption of accessing the large register file. Compared with the baseline and previous scalar architectures, G-Scalar improves power efficiency by 24% and 15%, respectively, at a negligible cost.

The cessation of Moore's Law has limited further improvements in power efficiency. In recent years, the physical realization of the memristor has demonstrated a promising solution to ultra-integrated hardware realization of neural networks, which can be leveraged for better performance and power efficiency gains. In this work, we introduce a power efficient framework for approximated computations by taking advantage of the memristor-based multilayer neural networks. A programmable memristor approximated computation unit (Memristor ACU) is introduced first to accelerate approximated computation and a memristor-based approximated computation framework with scalability is proposed on top of the Memristor ACU. We also introduce a parameter configuration algorithm of the Memristor ACU and a feedback state tuning circuit to program the Memristor ACU effectively. Our simulation results show that the maximum error of the Memristor ACU for 6 common complex functions is only 1.87% while the state tuning circuit can achieve 12-bit precision. The implementation of HMAX model atop our proposed memristor-based approximated computation framework demonstrates 22× power efficiency improvements than its pure digital implementation counterpart.

In recent years, optical neural networks have attracted widespread attention, due to their advantages of high speed, high parallelism, high bandwidth, and low power consumption. Photonic unitary neural network is a kind of neural networks that utilize the principles of unitary matrices and photonics to perform computations. In this paper, we design a photonic unitary neural network based on Mach–Zehnder interferometer arrays. The results show that the network has a good performance on both triangular and circular binary classification datasets, where most of the data points are correctly classified. The accuracies achieve 97% and 95% for triangular and circular datasets, with the loss function values of 0.023 and 0.046, respectively.

Event Abstract Back to Event Logarithmic Hybrid Optical Neural Network-type systems: towards digital-optical cognitive models of robust object recognition Ioannis Kypraios1* 1 Oxford University, Centre for Innovation & Enterprise, APEM Computing Lab, United Kingdom Abstract - By combining the complex logarithmic r-theta mapping of a space-variant imaging sensor [1, 2] with the hybrid digital-optical neural network filter together with a window unit multiple objects of the same class or of different classes can robustly be recognized. The resulted logarithmic r-theta mapping for hybrid digital-optical neural network system or briefly referred to as logarithmic hybrid optical neural network system is shown to exhibit with a single pass over the input data simultaneously in-plane rotation, out-of-plane rotation, scale, log r-θ map translation and shift invariance, and good clutter tolerance by recognizing correctly the different objects within the cluttered scenes. Here, we study the biologically-inspired knowledge learning and knowledge representation [3] achieved by the logarithmic hybrid optical neural network-type of object recognition systems (see Fig. 1 and Fig. 2). We investigate the effects that altering the knowledge representation can have on the problem’s learned knowledge and the problem solving process, in overall. Further, we study the logarithmic unconstrained-, logarithmic constrained-, and logarithmic modified-hybrid optical neural network systems architectures’ designs [4, 5]. We show how the logarithmic unconstrained-hybrid optical neural network object recognition system applies an unconstrained representation of the problem’s knowledge to maximize the search of solutions, how the logarithmic constrained-hybrid optical neural network object recognition system uses a constrained representation of the problem’s knowledge to guide the search towards certain solutions over others in the multidimensional search space, and how the logarithmic modified-hybrid optical neural network object recognition system uses memory-like masks to recall certain solutions over others in the multidimensional search space. Fig. 1. (a) Simplified human retina and visual cortex model used for description purposes only; (b) A digital-optical computational model for the cognitive interaction between the retina and the human visual cortex with the general logarithmic hybrid optical neural network architercture. Fig. 2. Biologically-inspired knowledge learning and knowledge representation with the logarithmic hybrid optical neural network-type of object recognition systems. Figure 1 Figure 2 References [1] C. F. R. Weiman, “Video compression via log-polar mapping”, Real-Time Image Processing II, SPIE Symposium on OE/Aerospace Sensing, Vol. 1295, 266-277 (1990) [2] C. G. Ho et al., “Sensor geometry and sampling methods for space-variant image processing”, Pattern Analysis and Applications, Vol. 5, 369-384 (2002) [3] I. Lee, and B. Portier, "An empirical study of knowledge representation and learning within conceptual spaces for intelligent agents", 6th IEEE/ACIS International Conference Computer and Information Science, 463-468, Melbourne, Australia (2007) [4] I. Kypraios et al., “Fully invariant complex Logarithmic r- map for the Hybrid Optical Neural Network filter for object recognition within cluttered scenes”, 50th Anniversary International Symposium ELMAR 2008, IEEE Region 8/IEEE Croatia Section/EURASIP, Zadar-Croatia, Vol. 1, 141-146 (2008) [5] I. Kypraios et al., “Logarithmic r-θ mapping for hybrid optical neural network filter for object recogntion within cluttered scenes”, Recent Advances in Multimedia Signal Processing and Communications, SCI231, Springer-Verlag Berlin Heidelberg, 91-120 (2009) Keywords: digital-optical, cognitive, object recognition, Robust and Adaptive Systems, Knowledge representation, learning and memory Conference: Neuroinformatics 2013, Stockholm, Sweden, 27 Aug - 29 Aug, 2013. Presentation Type: Poster Topic: Neuromorphic engineering Citation: Kypraios I (2013). Logarithmic Hybrid Optical Neural Network-type systems: towards digital-optical cognitive models of robust object recognition. Front. Neuroinform. Conference Abstract: Neuroinformatics 2013. doi: 10.3389/conf.fninf.2013.09.00066 Copyright: The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers. They are made available through the Frontiers publishing platform as a service to conference organizers and presenters. The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated. Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed. For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions. Received: 09 Apr 2013; Published Online: 11 Jul 2013. * Correspondence: Dr. Ioannis Kypraios, Oxford University, Centre for Innovation & Enterprise, APEM Computing Lab, Begbroke, Oxfordshire, OX5 1PF, United Kingdom, ioanniskyp@yahoo.com Login Required This action requires you to be registered with Frontiers and logged in. To register or login click here. Abstract Info Abstract The Authors in Frontiers Ioannis Kypraios Google Ioannis Kypraios Google Scholar Ioannis Kypraios PubMed Ioannis Kypraios Related Article in Frontiers Google Scholar PubMed Abstract Close Back to top Javascript is disabled. Please enable Javascript in your browser settings in order to see all the content on this page.

Approximate computing has emerged as a promising approach to energy-efficient design of digital systems in many domains such as digital signal processing, robotics, and machine learning. Numerous studies report that employing different data formats in Deep Neural Networks (DNNs), the dominant Machine Learning approach, could allow substantial improvements in power efficiency considering an acceptable quality for results. In this work, the application of Tunable Floating-Point (TFP) precision to DNN is presented. In TFP different precisions for different operations can be set by selecting a specific number of bits for significand and exponent in the floating-point representation. Flexibility in tuning the precision of given layers of the neural network may result in a more power efficient computation.

Optoelectronic realization of adaptive filters and equalizers using fiber optic tapped delay lines and spatial light modulators has been discussed recently. We describe the design of a single layer fiber optic Adaline neural network which can be used as a bit pattern classifier. In our realization we employ as few electronic devices as possible and use optical computation to utilize the advantages of optics in processing speed, parallelism, and interconnection. The new optical neural network described in this paper is designed for optical processing of guided lightwave signals, not electronic signals. We analyzed the convergence or learning characteristics of the optically implemented Adaline in the presence of errors in the hardware, and we studied methods for improving the convergence rate of the Adaline.© (1993) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

In contemporary academic settings, end-of-semester student feedback on a lecturer’s teaching abilities often fails to provide a comprehensive, real-time evaluation of their proficiency, and becomes less relevant with each new cohort of students. To address these limitations, an innovative feedback method has been proposed, utilizing image processing algorithms to dynamically assess the emotional states of students during lectures by analyzing their facial expressions. This real-time approach enables lecturers to promptly adapt and enhance their teaching techniques. Recognizing and engaging with emotionally positive students has been shown to foster better learning outcomes, as their enthusiasm actively stimulates cognitive engagement and information analysis. The purpose of this work is to identify emotions based on facial expressions using a deep learning model based on a convolutional neural network (CNN), where facial recognition is performed using the Viola–Jones algorithm on a group of students in a learning environment. The algorithm encompasses four key steps: image acquisition, preprocessing, emotion detection, and emotion recognition. The technological advancement of this research lies in the proposal to implement photonic hardware and create an optical neural network which offers unparalleled speed and efficiency in data processing. This approach demonstrates significant advancements over traditional electronic systems in handling computational tasks. An experimental validation was conducted in a classroom with 45 students, demonstrating that the level of understanding in the class as predicted was 43–62.94%, and the proposed CNN algorithm (facial expressions detection) achieved an impressive 83% accuracy in understanding students’ emotional states. The correlation between the CNN deep learning model and the students’ feedback was 91.7%. This novel approach opens avenues for the real-time assessment of students’ engagement levels and the effectiveness of the learning environment, providing valuable insights for ongoing improvements in teaching practices.

With advancements in science and technology, there has been phenomenal developments in the application of neural networks in dentistry. This systematic review aimed to report on the effectiveness of artificial intelligence (AI) applications designed for endodontic diagnosis, decision-making, and prediction of prognosis. Studies reporting on AI applications in endodontics were identified from the electronic databases such as PubMed, Medline, Embase, Cochrane, Google Scholar, Scopus, and Web of Science, for original research articles published from January 1, 2000, to June 1, 2020. A total of 10 studies that met our eligibility criteria were further analyzed for qualitative data. QUADAS-2 was applied for synthesis of the quality of the studies included. A wide range of AI applications have been implemented in endodontics. The neural networks employed were mostly based on convolutional neural networks (CNNs) and artificial neural networks (ANNs) in their neural architectures. These AI models have been used for locating apical foramen, retreatment predictions, prediction of periapical pathologies, detection and diagnosis of vertical root fractures, and assessment of root morphologies. These studies suggest that the neural networks performed similar to the experienced professionals in terms of accuracy and precision. In some studies, these models have even outperformed the specialists. These models can be of greater assistance as an expert opinion for less experienced and nonspecialists.

Optical neuromorphic computing has demonstrated promising performance with ultra-high computation speed, high bandwidth, and low energy consumption. The traditional optical neural network (ONN) architectures realize neuromorphic computing via electrical weight encoding. However, previous ONN design methodologies can only handle static linear projection with stationary synaptic weights, thus fail to support efficient and flexible computing when both operands are dynamically-encoded light signals. In this work, we propose a novel ONN engine O2NN based on wavelength-division multiplexing and differential detection to enable high-performance, robust, and versatile photonic neural computing with both light operands. Balanced optical weights and augmented quantization are introduced to enhance the representability and efficiency of our architecture. Static and dynamic variations are discussed in detail with a knowledge-distillation-based solution given for robustness improvement. Discussions on hardware cost and efficiency are provided for a comprehensive comparison with prior work. Simulation and experimental results show that the proposed ONN architecture provides flexible, efficient, and robust support for high-performance photonic neural computing with fully-optical operands under low-bit quantization and practical variations.

Machine learning with artificial neural networks has recently transformed many scientific fields by introducing new data analysis and information processing techniques. Despite these advancements, efficient implementation of machine learning on conventional computers remains challenging due to speed and power constraints. Optical computing schemes have quickly emerged as the leading candidate for replacing their electronic counterparts as the backbone for artificial neural networks. Some early integrated photonic neural network (IPNN) techniques have already been fast-tracked to industrial technologies. This review article focuses on the next generation of optical neural networks (ONNs), which can perform machine learning algorithms directly in free space. We have aptly named this class of neural network model the free space optical neural network (FSONN). We systematically compare FSONNs, IPNNs, and the traditional machine learning models with regard to their fundamental principles, forward propagation model, and training process. We survey several broad classes of FSONNs and categorize them based on the technology used in their hidden layers. These technologies include 3D printed layers, dielectric and plasmonic metasurface layers, and spatial light modulators. Finally, we summarize the current state of FSONN research and provide a roadmap for its future development.

While numerous artificial neural network (ANN) models have been electronically implemented and simulated by conventional computers, optical technology provides a far superior mechanism for the implementation of large-scale ANNs. The properties of light make it an ideal carrier of data signals. With optics, very large and high speed neural network architectures are possible. Because light is a predictable phenomenon, it can be described mathematically and its behavior can be simulated by conventional computers. A hologram is in essence a capture of the light field at a particular moment in time and space. Later, the hologram can be used to reconstruct the three dimensional light field carrying optical data. This makes a hologram an ideal medium for capturing, storing, and transmitting data in optical computers, such as optical neural networks (ONNs). Holograms can be created using conventional methods, but they can also be computer generated. In this paper, we will present an overview of optical neural networks, with emphasis on the holographic neural networks. We will take a look at the mathematical basis of holography in terms of the Fresnel Zone Plate and how it can be utilized in making computer generated holograms (CGHs). Finally, we will present various methods of CGH implementation in a two layer holographic ONN.