Photonic solutions are today a mature industrial reality concerning high speed, throughput data communication and switching infrastructures. It is still matter of investigation to what extent photonics will play role in next-generation computing architectures. In particular, due the recent outstanding achievements artificial neural networks, there big interest trying improve their speed energy efficiency by exploiting photonic-based hardware instead electronic-based hardware. this work we...

Abstract Linear programming is a very well known and deeply applied field of optimization theory. One its most famous used algorithms the so called Simplex algorithm, independently proposed by Kantorovič Dantzig, between end 30s 40s. Even if extremely powerful, algorithm suffers one initialization issue: starting point must be feasible basic solution problem to solve. To overcome it, two approaches may used: two-phases method Big-M method, both presenting positive negative aspects. In this...

This paper presents a flexible approach to forecasting of energy production in solar photovoltaic (PV) installations, using time series analysis and neural networks. Our goal is develop one day-ahead model based on an artificial network with tapped delay lines. Despite some methods already exist for problems, the main novelty our proposal tool technician PV installation correctly configure according particular characteristics. The correct configuration takes into account number hidden...

This paper presents a method, based on classification techniques, for automatic detection and diagnosis of defects rolling element bearings. The experimental data set consists vibration signals recorded by four accelerometers mechanical device including bearings: the were collected both with all faultless bearings after substituting one bearing an artificially damaged one. Four and, them, three severity levels are considered. Classification accuracy higher than 99% was achieved in...

Abstract Photonics-based neural networks promise to outperform electronic counterparts, accelerating network computations while reducing power consumption and footprint. However, these solutions suffer from physical layer constraints arising the underlying analog photonic hardware, impacting resolution of (in terms effective number bits), requiring use positive-valued inputs, imposing limitations in fan-in size convolutional kernels. To abstract constraints, this paper we introduce concept...

This article focuses on the trends, opportunities, and challenges of novel arithmetic for deep neural network (DNN) signal processing, with particular reference to assisted- autonomous driving applications. Due strict constraints in terms latency, dependability, security driving, machine perception (i.e., detection decision tasks) based DNNs cannot be implemented by relying remote cloud access. These tasks must performed real time embedded systems board vehicle, particularly inference phase...

Nowadays, two groundbreaking factors are emerging in neural networks. First, there is the RISC-V open instruction set architecture (ISA) that allows a seamless implementation of custom sets. Second, several novel formats for real number arithmetic. In this work, we combined these key aspects using very promising posit format, developing light Posit Processing Unit (PPU-light). We present an extension base ISA conversion between 8 or 16-bit posits and 32-bit IEEE Floats fixed point order to...

Reconfigurable linear optical processors can be used to perform transformations and are instrumental in effectively computing matrix–vector multiplications required each neural network layer. In this paper, we characterize compare two thermally tuned photonic integrated realized silicon-on-insulator silicon nitride platforms suited for extracting feature maps convolutional networks. The reduction bit resolution when crossing the processor is mainly due losses, range 2.3–3.3 chip 1.3–2.4...

This paper discusses the introduction of an integrated Posit Processing Unit (PPU) as alternative to Floating-point (FPU) for Deep Neural Networks (DNNs) in automotive applications. Autonomous Driving tasks are increasingly depending on DNNs. For example, detection obstacles by means object classification needs be performed real-time without involving remote computing. To speed up inference phase DNNs CPUs on-board vehicle should equipped with co-processors, such GPUs, which embed specific...

With increasing real-time constraints being put on the use of Deep Neural Networks (DNNs) by scenarios, there is need to review information representation. A very challenging path employ an encoding that allows a fast processing and hardware-friendly representation information. Among proposed alternatives IEEE 754 standard regarding floating point real numbers, recently introduced Posit format has been theoretically proven be really promising in satisfying mentioned requirements. However,...

This paper describes an optimal sampling approach to support glider fleet operators and marine scientists during the complex task of planning missions fleets underwater gliders. Optimal sampling, which has gained considerable attention in last decade, consists paths gliders minimize a specific criterion pertinent phenomenon under investigation. Different criteria (e.g., A, G, or E optimality), used geosciences obtain optimum design, lead different strategies. In particular, A produces for...

Prisoner's Dilemma (PD) is a widely studied game that plays an important role in Game Theory. This paper aims at extending PD Tournaments to the case of infinite, finite or infinitesimal payoffs using Sergeyev's Infinity Computing (IC). By exploiting IC, we are able show limits classical approach analysis theory, both sets feasible and numerically computable tournaments. In particular provide numerical computation exact outcome simple Tournament where one player meets every other infinite...

This article concerns the study of mixed Pareto-lexicographic multiobjective optimization problems where objectives must be partitioned in multiple priority levels (PLs). A PL is a group having same importance terms and subsequent decision making, while between PLs lexicographic ordering exists. naive approach would to define multilevel dominance relationship apply standard EMO/EMaO algorithm, but concept does not conform stable process as resulting violates transitive property needed...