In this article, we present a novel approach that seamlessly integrates machine learning (ML) algorithms into wearable technology through the use of weightless neural networks (WNNs) and flexible integrated circuits (FlexICs). Our methodology employs combinational intelligent networks (COIN) for edge inference on resource-constrained devices, highlighting the advantages of WNNs in terms of power efficiency and minimal hardware requirements. We propose an automated design flow for implementing COIN as FlexICs aimed at developing scalable, cost-effective, and environmentally sustainable wearable monitoring solutions. As a proof-of-concept demonstrator, an arrhythmia detection FlexIC was fabricated using COIN to meet the stringent requirements of medium-complexity wearable applications, offering a promising path toward personalized and accessible healthcare solutions.

Maritime Situational Awareness (MSA) is crucial for identifying suspicious vessel activities, such as dark-ship operations and prolonged loitering activities. However, the development of robust detection systems requires high-quality datasets that capture vessel encounters, particularly encounters that occur beyond 20 nautical miles (NM) from the coast. This paper presents the creation and analysis of a comprehensive data set that contains vessel trajectories associated with offshore encounters. The dataset, constructed using 12 months of data from the Marine Cadastre Automatic Identification System (AIS), leverages the H3 geohash system for spatial proximity detection and MovingPandas for trajectory extraction. The dataset analysis demonstrates that the dataset is a powerful tool for enhancing Maritime Domain Awareness (MDA), contributing to monitoring and security in the maritime environment. The analysis of encounter patterns highlights both the importance of reliable data and the need for a robust detection system to address uncertainties and information gaps.

Detecting fishing trajectories in maritime surveillance is of the utmost importance for identifying illegal fishing activity. In the event of illegal fishing activity, the maritime authority can mobilize resources to engage the vessel; hence, a false flag can be costly. This study investigates the efficacy of ensemble learning techniques for boosting individual model performance and decreasing uncertainty. Employing a range of machine learning models, including logistic regression, decision trees, random forests, neural networks, gradient boosting, and recurrent neural networks, the research evaluates the combination of these using ensemble methods like ensemble mean, weighted ensemble, and stacking approaches to enhance precision and decrease uncertainty. The primary dataset comprises a combination of fishing vessel and cargo vessel trajectories to train and test the models. Methodologically, the paper details the process of data analysis and the application of ensemble learning. A comparative assessment of individual models versus ensemble techniques forms the crux of this study. Results indicate a marked improvement in accuracy and consistency when employing ensemble methods, with weighted and stacking ensembles showing particular promise. These findings suggest that ensemble models outperform their individual counterparts in the context of maritime surveillance. This research makes a notable contribution to the maritime surveillance domain, demonstrating the potential of ensemble learning in enhancing detection capabilities for illegal fishing activities. The implications of these advancements are critical for maritime authorities as they strive to effectively monitor and protect marine ecosystems.

We introduce Combinational Intelligent Networks (COIN), a machine learning technique that targets edge inference using low-resourced FPGAs or ASICs. COIN is an improvement on LogicWiSARD, a recent weightless neural network that achieves low power, small area, and high throughput. We convert the LogicWiSARD model into a binary neural network, train it using backpropagation, and then convert it to a COIN model. As a result, COIN can achieve higher accuracy than LogicWiSARD or it can require significantly fewer hardware resources when comparing models with similar accuracies. In comparison to a BNN implementation, FINN, small and large COIN models are more energy efficient demonstrating up to 11.5x higher inferences/Joule at similar accuracy. Our tool executes the complete flow, from training to RTL. and is publicly available.

Algorithms for mobile networking are increasingly being moved from centralized servers towards the edge in order to decrease latency and improve the user experience. While much of this work is traditionally done using ASICs, 6G emphasizes the adaptability of algorithms for specific user scenarios, which motivates broader adoption of FPGAs. In this paper, we propose the FPGA-based Weightless Intrusion Warden (FWIW), a novel solution for detecting anomalous network traffic on edge devices. While prior work in this domain is based on conventional deep neural networks (DNNs), FWIW incorporates a weightless neural network (WNN), a table lookup-based model which learns sophisticated nonlinear behaviors. This allows FWIW to achieve accuracy far superior to prior FPGA-based work at a very small fraction of the model footprint, enabling deployment on small, low-cost devices. FWIW achieves a prediction accuracy of 98.5% on the UNSW-NB15 dataset with a total model parameter size of just 192 bytes, reducing error by 7.9x and model size by 262x vs. LogicNets, the best prior edge-optimized implementation. Implemented on a Xilinx Virtex UltraScale+ FPGA, FWIW demonstrates a 59x reduction in LUT usage with a 1.6x increase in throughput. The accuracy of FWIW comes within 0.6% of the best-reported result in literature (Edge-Detect), a model several orders of magnitude larger. Our results make it clear that WNNs are worth exploring in the emerging domain of edge networking, and suggest that FPGAs are capable of providing the extreme throughput needed.

‘‘Extreme edge” 1 devices, such as smart sensors, are a uniquely challenging environment for the deployment of machine learning. The tiny energy budgets of these devices lie beyond what is feasible for conventional deep neural networks, particularly in high-throughput scenarios, requiring us to rethink how we approach edge inference. In this work, we propose ULEEN, a model and FPGA-based accelerator architecture based on weightless neural networks (WNNs). WNNs eliminate energy-intensive arithmetic operations, instead using table lookups to perform computation, which makes them theoretically well-suited for edge inference. However, WNNs have historically suffered from poor accuracy and excessive memory usage. ULEEN incorporates algorithmic improvements and a novel training strategy inspired by binary neural networks (BNNs) to make significant strides in addressing these issues. We compare ULEEN against BNNs in software and hardware using the four MLPerf Tiny datasets and MNIST. Our FPGA implementations of ULEEN accomplish classification at 4.0–14.3 million inferences per second, improving area-normalized throughput by an average of 3.6× and steady-state energy efficiency by an average of 7.1× compared to the FPGA-based Xilinx FINN BNN inference platform. While ULEEN is not a universally applicable machine learning model, we demonstrate that it can be an excellent choice for certain applications in energy- and latency-critical edge environments.

Weightless neural networks (WNNs) are a class of machine learning model which use table lookups to perform inference, rather than the multiply-accumulate operations typical of deep neural networks (DNNs). Individual weightless neurons are capable of learning non-linear functions of their inputs, a theoretical advantage over the linear neurons in DNNs, yet state-of-the-art WNN architectures still lag behind DNNs in accuracy on common classification tasks. Additionally, many existing WNN architectures suffer from high memory requirements, hindering implementation. In this paper, we propose a novel WNN architecture, BTHOWeN, with key algorithmic and architectural improvements over prior work, namely counting Bloom filters, hardware-friendly hashing, and Gaussian-based nonlinear thermometer encodings. These enhancements improve model accuracy while reducing size and energy per inference. BTHOWeN targets the large and growing edge computing sector by providing superior latency and energy efficiency to both prior WNNs and comparable quantized DNNs. Compared to state-of-the-art WNNs across nine classification datasets, BTHOWeN on average reduces error by more than 40% and model size by more than 50%. We demonstrate the viability of a hardware implementation of BTHOWeN by presenting an FPGA-based inference accelerator, and compare its latency and resource usage against similarly accurate quantized DNN inference accelerators, including multi-layer perceptron (MLP) and convolutional models. The proposed BTHOWeN models consume almost 80% less energy than the MLP models, with nearly 85% reduction in latency. In our quest for efficient ML on the edge, WNNs are clearly deserving of additional attention.

Weightless neural networks (WNNs) are an alternative pattern recognition technique where RAM nodes function as neurons. As both training and inference require mostly table lookups, few additions, and no multiplications, WNNs are suitable for high-performance and low-power embedded applications. This work introduces a novel approach to implement WiSARD, the leading WNN state-of-the-art architecture, completely eliminating memories and arithmetic circuits and utilizing only logic functions. The approach creates compressed minimized implementations by converting trained WNN nodes from lookup tables to logic functions. The proposed LogicWiSARD is implemented in FPGA and ASIC technologies to illustrate its suitability for edge inference. Experimental results show more than 80% reduction in energy consumption when the proposed LogicWiSARD model is compared with a multilayer perceptron network (MLP) of equivalent accuracy. Compared to previous work on FPGA implementations for WNNs, convolutional neural networks, and binary neural networks, the energy savings of LogicWiSARD range between 32.2% and 99.6%.

The binary encoding of real valued inputs is a crucial part of Weightless Neural Networks.The Linear Thermometer and its variations are the most prominent methods to determine binary encoding for input data but, as they make assumptions about the input distribution, the resulting encoding is sub-optimal and possibly wasteful when the assumption is incorrect.We propose a new thermometer approach that doesn't require such assumptions.Our results show that it achieves similar or better accuracy when compared to a thermometer that correctly assumes the distribution, and accuracy gains up to 26.3% when other thermometer representations assume an unsound distribution.

Weightless neural networks (WNNs) are a type of machine learning model which perform prediction using lookup tables (LUTs) instead of arithmetic operations.Recent advancements in WNNs have reduced model sizes and improved accuracies, reducing the gap in accuracy with deep neural networks (DNNs).Modern DNNs leverage "pruning" techniques to reduce model size, but this has not previously been explored for WNNs.We propose a WNN pruning strategy based on identifying and culling the LUTs which contribute least to overall model accuracy.We demonstrate an average 40% reduction in model size with at most 1% reduction in accuracy.

Conditional branch prediction is a technique used to speculatively execute instructions before knowing the direction of conditional branch statements. Perceptron-based predictors have been extensively studied, however, they need large input sizes for the data to be linearly separable. To learn nonlinear functions from the inputs, we propose a conditional branch predictor based on the WiSARD model and compare it with two state-of-the-art predictors, the TAGE-SC-L and the Multiperspective Perceptron. We show that the WiSARD-based predictor with a smaller input size outperforms the perceptron-based predictor by about 0.09% and achieves similar accuracy to that of TAGE-SC-L.

While the 5G network technology is maturing and the number of commercial deployments is growing, the focus of the networking community is shifting to services and service delivery. 5G networks are designed to be a common platform for very distinct services with different characteristics. Network Slicing has been developed to offer service isolation between the different network offerings. Cloud-native services that are composed of a set of inter-dependent micro-services are assigned into their respective slices that usually span multiple service areas, network domains, and multiple data centers. Due to mobility events caused by moving end-users, slices with their assigned resources and services need to be re-scoped and re-provisioned. This leads to slice mobility whereby a slice moves between service areas and whereby the inter-dependent service and resources must be migrated to reduce system overhead and to ensure low-communication latency by following end-user mobility patterns. Recent advances in computational hardware, Artificial Intelligence, and Machine Learning have attracted interest within the communication community to study and experiment self-managed network slices. However, migrating a service instance of a slice remains an open and challenging process, given the needed co-ordination between inter-cloud resources, the dynamics, and constraints of inter-data center networks. For this purpose, we introduce a Deep Reinforcement Learning based agent that is using two different algorithms to optimize bandwidth allocations as well as to adjust the network usage to minimize slice migration overhead. We show that this approach results in significantly improved Quality of Experience. To validate our approach, we evaluate the agent under different configurations and in real-world settings and present the results.

Even a wider set of highly critical and latency-sensitive applications with resource needs from the access network and the edge will be supported by the 6G networks. Therefore, the 6G network will deal with diversification of service platforms. Optimizing the resource consumption of network slicing on top of a shared infrastructure will become essential to keep the operating costs on an acceptable level. Each vertical, e.g., eMBBPlus, BigCom, holographic and tactile communications can run on top of network slice with specific KPIs. Different verticals can have contradicting requirements running on top of the same infrastructure. This paper investigates the orchestration of network services within a federated end-to-end network slice, which may span over multiple cloud domains as expected to be a common scenario in 6G deployments. We introduce three optimization solutions that consider two conflicting objectives, the end-to-end delay and service relocation, for orchestrating network slice. While the first solution optimizes the end-to-end delay, the second solution optimizes the service relocation. Meanwhile, the third solution leverages the bargaining game theory for achieving optimal Pareto fair trade-off configuration to optimize both objectives. The simulation results demonstrate the efficiency of the proposed solutions to achieve their main design goals

The fifth generation of the mobile network aims to revolutionize mobile communication by offering both unparalleled performance and broader service offerings. 5G technology responds to the growing demand for higher bandwidth and lower latency, caused by a significant increase of connected resources. Leveraging on Software-Defined Networking (SDN) and artificial intelligence (AI) technologies, the 6G system can autonomously adapt to user requirements. This paper proposes a framework, named intelligent optimization framework (IoF), that leverages both network optimization and machine learning techniques for achieving the best performance results. The IoF framework allows for finding an exemplary resource allocation configuration of the mobile network by leveraging SDN technology. This work aims to configure the SDN-enabled switches and Open vSwitchs (OVSs) to enable an optimized data plane that reduces the overall operational expense (OPEX) cost and capital expense (CAPEX) cost within an optimized execution time. The evaluation results show our proposed framework's efficiency for delivering optimal configurations by reducing the number of allocated OVSs in a reasonable execution time.

Recent 5G trials have demonstrated the usefulness of the Network Slicing concept that delivers customizable services to new and under-serviced industry sectors. However, user mobility's impact on the optimal resource allocation within and between slices deserves more attention. Slices and their dedicated resources should be offered where the services are to be consumed to minimize network latency and associated overheads and costs. Different mobility patterns lead to different resource re-allocation triggers, leading eventually to slice mobility when enough resources are to be migrated. The selection of the proper triggers for resource re-allocation and related slice mobility patterns is challenging due to triggers' multiplicity and overlapping nature. In this paper, we investigate the applicability of two Deep Reinforcement Learning based algorithms for allowing a fine-grained selection of mobility triggers that may instantiate slice and resource mobility actions. While the first proposed algorithm relies on a value-based learning method, the second one exploits a hybrid approach to optimize the action selection process. We present an enhanced ETSI Network Function Virtualization edge computing architecture that incorporates the studied mechanisms to implement service and slice migration. We evaluate the proposed methods' efficiency in a simulated environment and compare their performance in terms of training stability, learning time, and scalability. Finally, we identify and quantify the applicability aspects of the respective approaches.

Among the existing computational parallel models, Gamma and Dynamic Dataflow are equivalent models where parallel programs can be developed in a natural way. However, the implementation of Gamma paradigm poses several communication and architectural challenges. On the other hand, interest-based protocols have emerged as possible solutions of communication routing enabling an efficient communication in IoT environments. An interesting property of Gamma is the possibility of locality exploration in a non-complex way that sounds quite suitable for IoT applications. This work proposes a novel execution model for Gamma programs integrating the Radnet --- an interest-based protocol --- as communication protocol. Since the interest processing using Gamma paradigm is not a trivial task, a dynamic dataflow graph is used to express the interest as edges between vertices. We explore the equivalence results between Gamma and dataflow and, also, provide experiments showing the potential of Gamma when used to implement approximate computing techniques.

Network Slicing (NS) is a key enabler of the upcoming 5G and beyond system, leveraging on both Network Function Virtualization (NFV) and Software Defined Networking (SDN), NS will enable a flexible deployment of Network Functions (NFs) belonging to multiple Service Function Chains (SFC) over various administrative and technological domains. Our novel architecture addresses the complexities and heterogeneities of verticals targeted by 5G systems, whereby each slice consists of a set of SFCs, and each SFC handles specific traffic within the slice. In this paper, we propose and evaluate a MILP optimization model to solve the complexities that arise from this new environment. Our proposed model enables a cost-optimal deployment of network slices allowing a mobile network operator to efficiently allocate the underlying layer resources according to its users' requirements. We also design a greedy-based heuristic to investigate the possible trade-offs between execution runtime and network slice deployment. For each network slice, the proposed solution guarantees the required delay and the bandwidth, while efficiently handling the use of both the VNF nodes and the physical nodes, reducing the service provider's Operating Expenditure (OPEX).

Software Defined Networking (SDN) is a driving technology for enabling the 5th Generation of mobile communication (5G) systems offering enhanced network management features and softwarization. This paper concentrates on reducing the operating expenditure (OPEX) costs while i) increasing the quality of service (QoS) by leveraging the benefits of queuing and multi-path forwarding in OpenFlow, ii) allowing an operator with an SDN-enabled network to efficiently allocate the network resources considering mobility, and iii) reducing or even eliminating the need for over-provisioning. For achieving these objectives, a QoS aware network configuration and multipath forwarding approach is introduced that efficiently manages the operation of SDN enabled open virtual switches (OVSs). This paper proposes and evaluates three solutions that exploit the strength of QoS aware routing using multiple paths. While the two first solutions provide optimal and approximate optimal configurations, respectively, using linear integer programming optimization, the third one is a heuristic that uses Dijkstra short-path algorithm. The obtained results demonstrate the performance of the proposed solutions in terms of OPEX and execution time.

Given the indispensable need for a reliable network architecture to cope with 5G networks, 3GPP introduced a covet technology dubbed 5G Service Based Architecture (5G-SBA). Meanwhile, Multi-access Edge Computing (MEC) combined with SBA conveys a better experience to end- users by bringing application hosting from centralized data centers down to the network edge, closer to consumers and the data generated by applications. Both the 3GPP and the ETSI proposals offered numerous benefits, particularly the ability to deliver highly customizable services. Nevertheless, compared to large data- centers that tolerate the hosting of standard virtualization technologies (Virtual Machines (VMs) and servers), MEC nodes are characterized by lower computational resources, thus the debut of lightweight micro-service based applications. Motivated by the deficiency of current micro-services-based applications to support users' mobility and assuming that all these issues are under the umbrella of Service Function Chain (SFC) migrations, we aim to introduce, explain and evaluate diverse SFC migration patterns. The obtained results demonstrate that there is no clear vanquisher, but selecting the right SFC migration pattern depends on users' motion, applications' requirements, and MEC nodes' resources.

The use cases of the upcoming 5G mobile networks introduce new and complex user demands that will require support for fast reconfiguration of network resources. Software Defined Network (SDN) is a key technology that can address these requirements, as it decouples the control plane from the data plane of the network devices and logically centralizes the control plane in the SDN controller. SDN network operating system (ONOS) is a state-of-art SDN controller that aims to address this important scalability limitation from its design. An important feature of ONOS is that it allows network administrators to configure and manage networks with a high-level of abstraction by using Intent specifications. An Intent is a policy expression describing what is the desired outcome rather than how the outcome should be reached. The concept of Intents coupled with the distributed storage space are the key components for the theoretical scalability of ONOS. In this paper, we present our evaluation of the ONOS Intent northbound interface using a methodology that takes into consideration the interface access method, type of Intent and number of installed Intents. Our preliminary analysis indicates a linear increase in the computational cost with regards to the number of submitted Intents, with the access method being a major factor in the overall computational cost.

Given the constantly growing demand for inter- data-center services that 5G networks are bringing, live migration has become a covet and very challenging technology. Meanwhile, the emergence of Software Defined Networking (SDN) and Network Function Virtualization (NFV) technologies has completely transformed modern networks by offering more flexibility and at the same time more complexity. So far, investigations have been confined to integrating the live migration process with SDN/NFV paradigms in order to ensure the desired Quality of Experience (QoE). However, the simple integration is not sufficient to handle unexpected cases such as resources' unavailability, networking issues, and system control. For this purpose, we present MIRA!, a novel framework for managing reliable live migrations of virtual resources across different Infrastructure as a Service (IaaS), handling unexpected cases, while ensuring high QoS and a very low downtime without human intervention using an SDN aware solution. To validate our proposed framework, we performed a set of experimental evaluations under different configurations. The obtained results of our proposed framework show a 21% time reduction compared to a prior work and an interesting behavior while modifying the number of allocated CPU cores.

Network Slicing (NS) is expected to be a key functionality of the upcoming 5G systems. Coupled with Software Defined Networking (SDN) and Network Function Virtualization (NFV), NS will enable a flexible deployment of Network Functions belonging to multiple Service Function Chains (SFC) over a shared infrastructure. To address the complexities that arise from this new environment, we formulate a MILP optimization model that enables a cost- optimal deployment of network slices, allowing a Mobile Network Operator to efficiently allocate the underlying layer resources according to the users' requirements. For each network slice, the proposed solution guarantees the required delay and the bandwidth, while efficiently handling the usage of underlying nodes, which leads to reduced cost. The obtained results show the efficiency of the proposed solution in terms of cost and execution time for small-scale networks, while it shows an interesting behavior in the optimization of the mapping of slices into underlay nodes of the large-scale topologies.

The future 5G systems ought to meet diverse requirements of new industry verticals, such as Massive Internet of Things (IoT), broadband access in dense networks and ultra-reliable communications. Network slicing is an important concept that is expected to support these 5G verticals and cope with the conflicting requirements of their respective services. Network slicing allows the deployment of multiple virtual networks, or slices, over the same physical infrastructure as well as supporting on-demand resource allocation to those slices. In this paper, we propose an architecture that will explore how both Network Function Virtualization (NFV) and Software Defined Networking (SDN) may be leveraged to secure a network slice on-demand, addressing the new security concerns imposed to the network management by the flexibility and elasticity support. Our proposed framework aims to ensure an optimal resource allocation that manages the slice security strategy in an efficient way. Moreover, experimental performance evaluations are presented to evaluate the security overhead in virtualized environments.

Software Defined Networking (SDN) is an emerging technology that will play an important role in enabling 5G, since it offers enhanced network management features. SDN allows programmability of the control plane, abstracting the underlying network infrastructure for applications and network services, e.g. through the OpenFlow protocol. In this paper, we propose a solution that enables the end-to-end Quality of Service (QoS) based on the queue support in OpenFlow, allowing an operator with a SDN-enabled network to efficiently allocate the network resources according to the users' demands, reducing or even eliminating the need for over-provisioning. For each traffic flow, the proposed solution guarantees the required end-to- end QoS, while efficiently managing the utilization of open virtual switches (OVSs), which leads to reduced cost. The cost could be also reduced as a fewer number of OVSs are needed, which are enabled in different data centers. For ensuring these objectives, the proposed solution explores the strength of multi-path routing based on SDN with a precise bandwidth allocation. The obtained results show the efficiency of the proposed solution in terms of cost and execution time.

Summary The use of Dynamic Voltage and Frequency Scaling (DVFS) by Energy‐Efficient (EE) computer systems considerably increases the requirements regarding the design of efficient system clocks. On the one hand, the operation of a system clock must support the independent operating frequencies of the processor core units, the dynamic migration of the running processes between the processors core units, and the use of synchronization and time interpolation techniques to maintain the accuracy of the system clock. On the other hand, an efficient system clock has to minimize the overhead of its own operation, aiming at energy efficiency of EE computer systems. In this paper, we present the design and evaluation of the RVEC virtual system clock for the EE Wireless Raspberry Pi (RasPi) platform. In the RasPi platform, the use of DVFS for reducing the energy consumption hinders the direct use of the cycle count of the ARM11 processor core for building an efficient system clock. Therefore, a distinct feature of RVEC is to obviate this obstacle, such that it can make use of the cycle count circuit for precise and accurate time measurements, concurrently with the use of DVFS by the operating system of the ARM11 processor core. Specifically, this paper presents the design and experimental evaluation of an implementation of the RVEC virtual system clock in the Linux kernel of the RasPi platform with DVFS. Our experimental results validate the RVEC virtual system clock as an efficient system clock for the EE RasPi platform that runs the Linux operating system. Copyright © 2016 John Wiley & Sons, Ltd.

Energy-efficient computer systems are making increasing use of processors that have multiple core units, DVFS, and virtualization support. However, current system clocks have not been usually designed to cope with the capacity of such mechanisms to decelerate/accelerate the passage of time, which increases the time drifts in the system and produces two adverse side effects. First, a reduction in the precision of the system clocks, which makes it infeasible to run applications that are dependent on precise time measurements. Second, increasing the rate of system resynchronization with an external global clock, which adds more noise to the system and counteracts the attainment of a desirable energy efficiency. As an alternative to the system clock, we propose an original virtual clock, named RVEC, with the property that the time count is strictly increasing and precise (SIP). A preliminary experimental evaluation of an implementation of RVEC in Linux using a beowulf cluster of four energy-efficient computer systems showed that RVEC exhibited the SIP property while was highly precise and had negligible overhead in comparison with representative Linux system clocks. Furthermore, we used RVEC to build a High-Precision Global Clock (HPGC) which is free from resynchronization and implemented HPGC in the OpenMPI library as a time synchronization service for the MPI_Wtime() function to improve its timekeeping functions and lower the system noise. Our preliminary results from micro benchmarks executing in the same cluster indicated that the HPGC is highly scalable and precise solution which allowed the micro benchmarks to stay globally synchronized by using only 30 messages per node to initially synchronize the cluster nodes, thanks to the RVEC's SIP property. These results suggest that RVEC and HPGC can be effective alternatives to the system clock and the global clock respectively, in energy-efficient computer systems, especially for MPI applications running on beowulf clusters.