Flexible Architecture Research Articles

An open-source platform for sub-textrm{g}, sub-upmuA data loggers

BackgroundRapid improvements in inexpensive, low-power, movement and environmental sensors have sparked a revolution in animal behavior research by enabling the creation of data loggers (henceforth, tags) that can capture fine-grained behavioral data over many months. Nevertheless, development of tags that are suitable for use with small species, for example, birds under 25 g, remains challenging because of the extreme mass (under 1textrm{g}) and power (average current under 1upmuA) constraints. These constraints dictate that a tag should carry exactly the sensors required for a given experiment and the data collection protocol should be specialized to the experiment. Furthermore, it can be extremely challenging to design hardware and software to achieve the energy efficiency required for long tag life.ResultsWe present an activity monitor, BitTag, that can continuously collect activity data for 4–12 months at 0.5–0.8textrm{g}, depending upon battery choice, and which has been used to collect more than 500,000 h of data in a variety of experiments. The BitTag architecture provides a general platform to support the development and deployment of custom sub-textrm{g} tags. This platform consists of a flexible tag architecture, software for both tags and host computers, and hardware to provide the host/tag interface necessary for preparing tags for “flight” and for accessing tag data “post-flight”. We demonstrate how the BitTag platform can be extended to quickly develop novel tags with other sensors while satisfying the 1g/1upmuA mass and power requirements through the design of a novel barometric pressure sensing tag that can collect pressure and temperature data every 60textrm{s} for a year with mass under 0.6textrm{g}.

High-Performance of Eigenvalue Decomposition on FPGA for the DOA Estimation

For the direction of arrival (DOA) in array signal processing, eigenvalue decomposition (EVD) is one key issue in hardware implementation of the multiple signal classification (MUSIC) algorithm. Therefore, we introduce the look-ahead sim- plified one-sided Jacobi's method to efficiently decompose those symmetric matrices in this article and prove that the new method has the best orthogonality of eigenvector and locates eigenvectors closest to the true solution in theory. Both the numerical perform- ance and real-time are important in engineering, so we present the novel flexible hardware architecture in single floating point arithmetic for EVD on field-programmable gate arrays (FPGAs). Finally, the simulated and raw data are used to investigate the performance of some different approaches in the context of both the EVD and MUSIC algorithm. The experimental results show that our proposed method has the best performance.

HALO: A Hardware–Software Co-Designed Processor for Brain–Computer Interfaces

Brain-computer interfaces (BCIs) enable direct communication with the brain, providing valuable information about brain function and enabling novel treatment of brain disorders. Our group has been building HALO , a flexible and ultra-lowpower processing architecture for BCIs. HALO can process up to 46 Mbps of neural data, a significant increase over the interfacing bandwidth achievable by prior BCIs. HALO can also be programmed to support several applications, unlike most prior BCIs. Key to HALO ’s effectiveness is a hardware accelerator cluster, where each accelerator operates within its own clock domain. A configurable interconnect connects the accelerators to create data flow pipelines that realize neural signal processing algorithms. We have taped out our design in a 12nm CMOS process. The resulting chip runs at 0.88V, peraccelerator frequencies of 3–180 MHz, and consumes at most 5.0mW for each signal processing pipeline. Evaluations using electrophysiological data collected from a non-human primate confirm HALO ’s flexibility and superior performance per watt.

Hyper-convolutions via implicit kernels for medical image analysis

The convolutional neural network (CNN) is one of the most commonly used architectures for computer vision tasks. The key building block of a CNN is the convolutional kernel that aggregates information from the pixel neighborhood and shares weights across all pixels. A standard CNN's capacity, and thus its performance, is directly related to the number of learnable kernel weights, which is determined by the number of channels and the kernel size (support). In this paper, we present the hyper-convolution, a novel building block that implicitly encodes the convolutional kernel using spatial coordinates. Unlike a regular convolutional kernel, whose weights are independently learned, hyper-convolution kernel weights are correlated through an encoder that maps spatial coordinates to their corresponding values. Hyper-convolutions decouple kernel size from the total number of learnable parameters, enabling a more flexible architecture design. We demonstrate in our experiments that replacing regular convolutions with hyper-convolutions can improve performance with less parameters, and increase robustness against noise. We provide our code here: https://github.com/tym002/Hyper-Convolution.

Detection Optimization of Rare Attacks in Software-Defined Network using Ensemble Learning

Software-defined networking (SDN) is a highly flexible architecture that automates and facilitates network configuration and management. Intrusion detection systems (IDS) are becoming essential components in the network to detect malicious attacks and suspicious activities by continuously monitoring network traffic. Integration between SDN and machine learning (ML) techniques is extensively used to build an effective IDS against all potential cyber-attacks that aim at breaking the network security policy and stealing valuable data. Implementing an IDS based on SDN and ML has the advantage of managing traffic dynamically and fully autonomously to provide high protection against security threats. The main objective of this paper is to propose an IDS based on SDN that integrates adaptive boosting and cost-sensitive techniques to improve the detection rates of rare attacks without compromising the detection accuracy of familiar attacks. Cost matrix values for various attacks are optimized using grid search and genetic algorithms. The integration of AdaBoost and cost-sensitive aims to build a classification model that minimizes the total number of high-cost errors caused by incorrectly classifying attacks. Experiments were conducted on the CSE-CIC-IDS2017 and CSE-CIC-IDS2018 datasets to simulate the network. The proposed algorithm's performance is evaluated using the macro F1-score and the geometric average of the recall (G-mean). Simulation results prove that the proposed algorithm enhances the detection performance of rare intrusions compared to current techniques, where it achieves the macro F1-score of 0.99 and the G-mean of 0.992 using CIC-IDSC-IDS2017, while for CIC-IDSC-IDS2018 the macro F1-score is 0.993 and the G-mean is 0.994.

Reproducible Tract Profiles 2 (RTP2) suite, from diffusion MRI acquisition to clinical practice and research

Diffusion MRI is a complex technique, where new discoveries and implementations occur at a fast pace. The expertise needed for data analyses and accurate and reproducible results is increasingly demanding and requires multidisciplinary collaborations. In the present work we introduce Reproducible Tract Profiles 2 (RTP2), a set of flexible and automated methods to analyze anatomical MRI and diffusion weighted imaging (DWI) data for reproducible tractography. RTP2 reads structural MRI data and processes them through a succession of serialized containerized analyses. We describe the DWI algorithms used to identify white-matter tracts and their summary metrics, the flexible architecture of the platform, and the tools to programmatically access and control the computations. The combination of these three components provides an easy-to-use automatized tool developed and tested over 20 years, to obtain usable and reliable state-of-the-art diffusion metrics at the individual and group levels for basic research and clinical practice.

A General Convolutional Neural Network to Reconstruct Remotely Sensed Chlorophyll-a Concentration

Satellite-observed chlorophyll-a (Chl-a) concentrations are key to studies of phytoplankton dynamics. However, there are gaps in remotely sensed images mainly due to cloud coverage which requires reconstruction. This study proposed a method to build a general convolutional neural network (CNN) model that can reconstruct images in unfamiliar areas. Although several CNN models to reconstruct Chl-a in a specific area have already been proposed, the model in this research has the advantage of generality. The model uses a more flexible U-net architecture so that it can accept input of different shapes. Images from three areas of different shapes were used in model training to improve the generality of the model. Six models, with different auxiliary input schemes and architectures, were trained and evaluated. Results show that the model with bathymetry input and coarse-to-fine architecture has the best performance and can give reasonable reconstruction for the unfamiliar area. The best model shows better results than traditional interpolation methods when reconstructing for an unfamiliar area, especially in regions outside the data coverage.

UltraSEQ, a Universal Bioinformatic Platform for Information-Based Clinical Metagenomics and Beyond.

Applied metagenomics is a powerful emerging capability enabling the untargeted detection of pathogens, and its application in clinical diagnostics promises to alleviate the limitations of current targeted assays. While metagenomics offers a hypothesis-free approach to identify any pathogen, including unculturable and potentially novel pathogens, its application in clinical diagnostics has so far been limited by workflow-specific requirements, computational constraints, and lengthy expert review requirements. To address these challenges, we developed UltraSEQ, a first-of-its-kind accurate and scalable metagenomic bioinformatic tool for potential clinical diagnostics and biosurveillance utility. Here, we present the results of the evaluation of our novel UltraSEQ pipeline using an in silico-synthesized metagenome, mock microbial community data sets, and publicly available clinical data sets from samples of different infection types, including both short-read and long-read sequencing data. Our results show that UltraSEQ successfully detected all expected species across the tree of life in the in silico sample and detected all 10 bacterial and fungal species in the mock microbial community data set. For clinical data sets, even without requiring data set-specific configuration setting changes, background sample subtraction, or prior sample information, UltraSEQ achieved an overall accuracy of 91%. Furthermore, as an initial demonstration with a limited patient sample set, we show UltraSEQ's ability to provide antibiotic resistance and virulence factor genotypes that are consistent with phenotypic results. Taken together, the above-described results demonstrate that the UltraSEQ platform offers a transformative approach for microbial and metagenomic sample characterization, employing a biologically informed detection logic, deep metadata, and a flexible system architecture for the classification and characterization of taxonomic origin, gene function, and user-defined functions, including disease-causing infections. IMPORTANCE Traditional clinical microbiology-based diagnostic tests rely on targeted methods that can detect only one to a few preselected organisms or slow, culture-based methods. Although widely used today, these methods have several limitations, resulting in rates of cases of an unknown etiology of infection of >50% for several disease types. Massive developments in sequencing technologies have made it possible to apply metagenomic methods to clinical diagnostics, but current offerings are limited to a specific disease type or sequencer workflow and/or require laboratory-specific controls. The limitations associated with current clinical metagenomic offerings result from the fact that the backend bioinformatic pipelines are optimized for the specific parameters described above, resulting in an excess of unmaintained, redundant, and niche tools that lack standardization and explainable outputs. In this paper, we demonstrate that UltraSEQ uses a novel, information-based approach that enables accurate, evidence-based predictions for diagnosis as well as the functional characterization of a sample.

DIVINE: A pricing mechanism for outsourcing data classification service in data market

Although data mining plays an essential role in enhancing business decision-making, currently, there are hardly any suitable online platforms available that can facilitate outsourcing data classification services in data markets. This paper presents an efficient and flexible online service platform architecture for data classification and delves into an in-depth study on pricing design to maximize revenue. To construct this online service platform, we address three significant challenges. Firstly, we focus on efficiently modeling accurate classifiers while minimizing consumption. Secondly, we design a flexible pricing mechanism with multiple levels. Lastly, we maximize online revenue based on incomplete information. Through comprehensive consideration of these three challenges, this paper proposes a query-based Data classIfication serVice with onlIne priciNg mEchanism, named DIVINE, which determines the transaction price of data classification services by learning buyers' valuations online. The theoretical analysis of this paper illustrates that DIVINE can ensure a constant (1+ϵ) competitive ratio compared to the maximal offline revenue with complete information. Finally, DIVINE is evaluated on UCI databases. The experimental results show that it not only exhibits low consumption, multiple levels, and two kinds of robustness but also outperforms existing state-of-the-art pricing mechanisms in terms of revenue, which can achieve about 90% of the optimal offline revenue and validates the theoretical analysis.

Green Community Center: Using Long-life Material for Low Maintenance and Serviceability.

A community center is a building that unites various functions and is adapted to the character of the area and the needs of its residents. If it is related to the Sustainable Development Goals, then one of the targets can be met by providing public spaces that are safe, inclusive, and easily accessible to everyone. Specific space requirements and capacities of different users are a problem in community center so a movable partition system could be one solution. In addition, space configuration can also be arranged so that it is not monotonous by utilizing existing partitions. The problem in this research is how to choose a durable movable partition material for use in a community center. In addition, this study aims to determine durable movable partition materials so that the community center can not only meet the needs of its residents but also have flexible spaces and support sustainable architecture. The method used in this study is a comparative method by analyzing the durability of several materials commonly used as partition materials. Based on the results of the analysis and discussion of research data, the movable partition material that can be used is a PVC board. This is because PVC boards are resistant to water and termites and are not harmful to the environment.

STfusion: Fast and Flexible Multi-NN Execution Using Spatio-Temporal Block Fusion and Memory Management

To maximize the cost-effectiveness of neural network (NN) accelerators, architects are actively developing single-chip accelerators which can execute many NNs simultaneously. However, previous approaches fail to achieve full performance potential by exploiting only spatial or temporal resource sharing (SS or TS). They also do not consider memory management that can significantly affect performance. This limitation leads to the dire need for a new multi-NN accelerator taking both opportunities with careful memory management. But, it is extremely challenging to design an ideal spatio-temporal sharing accelerator because it requires (1) an algorithm that determines the degree of SS/TS in large exploration spaces, (2) a new STS-enabled accelerator devised with diverse design points, and (3) carefully-designed memory management that minimizes resource contention during numerous data transfers upon reconfiguration. To this end, we propose STfusion, a fast and flexible multi-NN execution architecture. First, STfusion partitions an accelerator into multiple smaller TS-enabled accelerators. Second, STfusion dynamically fuses small accelerators to adjust the accelerator sizes. Third, STfusion manages on-chip buffer in a page-granularity for stall-free data transfers. Lastly, STfusion provides an algorithm that determines the degree of SS/TS to achieve high throughput while satisfying QoS goals. Our evaluation shows that STfusion significantly outperforms state-of-the-art multi-NN accelerators.

Enabling Next-Generation Industrial Networks with Industrial PON

Traditional industrial networks primarily focus on onsite device operation and worker safety. They do not engage in the ongoing digital transformation with remote control machinery and big data processing. The next-generation of industrial networks is expected to offer high-speed connectivity, intelligent management, and flexible architecture to adapt to technology advances and business competitions. Data from the environment, machinery, and offices should be collected and analyzed to support automated procedures related to an industry. In this article, a network architecture based on the passive optical network (PON) technology, the so-called industrial PON, is proposed to enable the next-generation industrial networks. It manages the network resource via a three-layer structure. Recent industrial PON implementations deployed by a Tier 1 operator are investigated in detail to evaluate the solution feasibility and network performance. Results demonstrate that the industrial PON can connect various business assets, provide data exchange to automate industrial processes, and manage the business to improve production efficiency.

Smart home anomaly-based IDS: Architecture proposal and case study

The complexity and diversity of the technologies involved in the Internet of Things (IoT) challenge the generalization of security solutions based on anomaly detection, which should fit the particularities of each context and deployment and allow for performance comparison.In this work, we provide a flexible architecture based on building blocks suited for detecting anomalies in the network traffic and the application-layer data exchanged by IoT devices in the context of Smart Home. Following this architecture, we have defined a particular Intrusion Detector System (IDS) for a case study that uses a public dataset with the electrical consumption of 21 home devices over one year. In particular, we have defined ten Indicators of Compromise (IoC) to detect network attacks and two anomaly detectors to detect false command or data injection attacks. We have also included a signature-based IDS (Snort) to extend the detection range to known attacks. We have reproduced eight network attacks (e.g., DoS, scanning) and four False Command or Data Injection attacks to test our IDS performance. The results show that all attacks were successfully detected by our IoCs and anomaly detectors with a false positive rate lower than 0.3%. Signature detection was able to detect only 4 out of 12 attacks. Our architecture and the IDS developed can be a reference for developing future IDS suited to different contexts or use cases. Given that we use a public dataset, our contribution can also serve as a baseline for comparison with new techniques that improve detection performance.

Using a Flexible Model to Compare the Efficacy of Geographical and Temporal Contextual Information of Location-Based Social Network Data for Location Prediction

In recent years, next location prediction has been of paramount importance for a wide range of location-based social network (LBSN) services. The influence of geographical and temporal contextual information (GTCI) is crucial for analyzing individual behaviors for personalized point-of-interest (POI) recommendations. A number of studies have considered GTCI to improve the performance of POI prediction algorithms, but they have limitations. Moreover, reviewing the related literature revealed that no research has investigated and evaluated the GTCI of LBSN data for location prediction in the form presented in this study. Here, we extended the gated recurrent unit (GRU) model by adding additional attention gates to separately consider GTCI for location prediction based on LBSN data and introduced the extended attention GRU (EAGRU) model. Furthermore, we used the flexibility of the EAGRU architecture and developed it in four states to compare the efficacy of GTCI for location prediction for LBSN users. Real-world, large-scale datasets based on two LBSNs (Gowalla and Foursquare) were used for a complete review. The results revealed that the performance of the EAGRU model was higher than that of competitive baseline methods. In addition, the efficacy of the geographical CI was significantly higher than the temporal CI.

A robust‐link controller placement model for large‐scale software defined networks

AbstractThe software defined network (SDN) is a novel flexible network architecture. For the purpose of facilitating management, multiple controllers need to be placed in large‐scale SDN. When a link failure occurs, the path can be re‐planned by the control layer. However, unreasonable controller placement scheme brings about a sharp latency increase between controllers and switches. The reliability of the SDN is therefore affected. To this end, we define a link robust distance (LRD) by considering the states during link failures and non‐link failures. Then, we use it design a robust link controller placement model (RLCPM) which can reduce the impact of link failures on the delay between the controller and the switch. In addition, we design a heuristic algorithm with low‐time complexity to solve the RLCPM more effectively. Experimental results show that, compared with other models, RLCPM simplifies the controller placement problem when considering a link failure. The maximum latency between controllers and switches of the obtained controller placement scheme is lower.

Multi-connectivity in 5G New Radio: Optimal resource allocation for split bearer and data duplication

Mobile radio networks have been evolving towards the integration of services and devices with a diverse set of throughput, latency, and reliability requirements. To support these requirements, 3GPP has introduced Multi Connectivity (MC) as a more flexible architecture for 5G New Radio (NR), where multiple radio links can be simultaneously activated to split or duplicate data traffic. Multi connectivity improves single user performance at the cost of higher interference due to the increase of radio transmissions, which negatively affects system throughput.This paper analyzes the problem of admission control and resource allocation in multi connectivity scenarios, considering different requirements and 5G NR features. Specifically, we formulate two optimization problems that leverage the features of the Packet Data Convergence Protocol (PDCP) layer, which controls the flow of data packets of the data radio bearer: the PDCP Split-Bearer Decision (PSD) and the PDCP Duplication Decision (PDD) problems, which are tailored for the enhanced Mobile Broadband (eMBB) and Ultra Reliable Low Latency Communications (uRLLC) services, respectively. We further provide heuristic approaches, specifically designed for the PSD and PDD problems, to effectively solve both these problems.Numerical results in realistic network deployments confirm that our solutions can effectively allocate radio resources increasing admission rate and system throughput, while guaranteeing the required reliability level.

Design of a nonvolatile-register-embedded RISC-V CPU with software-controlled data-retention and hardware-acceleration functions

This paper describes the design of a nonvolatile CPU based on RISC-V that is an open-source and highly flexible instruction set architecture. This CPU incorporates nonvolatile registers utilizing magnetic tunnel junction (MTJ) device, as well as custom instructions specific to the control of these nonvolatile registers and an accelerator module embedded into the CPU architecture. These techniques enable efficient execution of intermittent operations suitable for energy-limited internet-of-things (IoT) applications. Through performance evaluation of the CPU designed in a 55-nm CMOS/MTJ-hybrid process technology, we show that our CPU can save up to 56.9% of power consumption compared to conventional ones, with an average power consumption of 3.91 μW/MHz.

SGAM-Based Analysis for the Capacity Optimization of Smart Grids Utilizing e-Mobility: The Use Case of Booking a Charge Session

The description of the functionality of a smart grid’s architectural concept, analyzing different Smart Grid (SG) scenarios without disrupting the smooth operation of the individual processes, is a major challenge. The field of smart energy grids has been increasing in complexity since there are many stakeholder entities with diverse roles. Electric Vehicles (EVs) can transform the stress on the energy grid into an opportunity to act as a flexible asset. Smart charging through an external control system can have benefits for the energy sector, both in grid management and environmental terms. A suitable model for analyzing and visualizing smart grid use cases in a technology-neutral manner is required. This paper presents a flexible architecture for the potential implementation of electromobility as a distributed storage asset for the grid’s capacity optimization by applying the Use Case and Smart Grid Architecture Model (SGAM) methodologies. The use case scenario of booking a charge session through a mobile application, as part of the TwinERGY Horizon 2020 project, is deployed to structure the SGAM framework layers and investigate the applicability of the SGAM framework in the integration of electromobility as a distributed storage asset into electricity grids with the objective of enhanced flexibility and decarbonization.

Towards High-Performance Supersingular Isogeny Cryptographic Hardware Accelerator Design

Cryptosystems based on supersingular isogeny are a novel tool in post-quantum cryptography. One compelling characteristic is their concise keys and ciphertexts. However, the performance of supersingular isogeny computation is currently worse than that of other schemes. This is primarily due to the following factors. Firstly, the underlying field is a quadratic extension of the finite field, resulting in higher computational complexity. Secondly, the strategy for large-degree isogeny evaluation is complex and dependent on the elementary arithmetic units employed. Thirdly, adapting the same hardware to different parameters is challenging. Considering the evolution of similar curve-based cryptosystems, we believe proper algorithm optimization and hardware acceleration will reduce its speed overhead. This paper describes a high-performance and flexible hardware architecture that accelerates isogeny computation. Specifically, we optimize the design by creating a dedicated quadratic Montgomery multiplier and an efficient scheduling strategy that are suitable for supersingular isogeny. The multiplier operates on Fp2 under projective coordinate formulas, and the scheduling is tailored to it. By exploiting additional parallelism through replicated multipliers and concurrent isogeny subroutines, our 65 nm SMIC technology cryptographic accelerator can generate ephemeral public keys in 2.40 ms for Alice and 2.79 ms for Bob with a 751-bit prime setting. Sharing the secret key costs another 2.04 ms and 2.35 ms, respectively.

Flexible Nanoarchitectonics for Biosensing and Physiological Monitoring Applications (Small 9/2023)

Flexible Mesoporous Electronics In article number 2204946, Hoang-Phuong Phan, Yusuke Yamauchi, Tuan-Khoa Nguyen, Mostafa Kamal Masud, and co-workers develop flexible mesoporous architectures using selective electrochemical deposition. Soft metallic nanopores formed on thin polyimide films offer excellent mechanical bendability, high sensitivity for glucose sensing, and low impedance that facilitates detection of neural activities. These features suggest the promise of soft mesoporous material systems for biomedical applications.