Ad-hoc Architecture Research Articles

Computing and evaluating saliency maps for image classification: a tutorial

Facing the black-box nature of deep learning models for image classification, a popular trend in the literature proposes methods to generate explanations in the form of heat maps indicating the areas that played an important role in the models’ decisions. Such explanations are called saliency maps and constitute an active field of research, given that many fundamental questions are yet to be answered: how to compute them efficiently? How to evaluate them? What exactly can they be used for? Given the increasing rate at which papers are produced and the vast amount of literature that is already existing, we propose our study to help newcomers become part of this community and to contribute to the research field. First, the two existing approaches to generate saliency maps are discussed, namely post-hoc methods and attention models. Post-hoc methods are generic algorithms that can be applied to any model from a given class without requiring fine-tuning. On the contrary, attention models are ad-hoc architectures that generate a saliency map during the inference phase to guide the decision. We show that both approaches can be divided into several subcategories and illustrate each of them with one important model or method. Second, we present the current methodologies used to evaluate saliency maps, including objective and subjective protocols, depending on whether or not they involve users. Among objective methods, we notably detail faithfulness metrics and propose an implementation featuring the faithfulness metrics discussed in this paper (https://github.com/TristanGomez44/metrics-saliency-maps).

A Flexible, High-Voltage (>100V) Generating Device Based on Zebra-Like Asymmetrical Photovoltaic Cascade.

The mutual conversion between light and electricity lies at the heart of optoelectronic and photonic applications. Maximization of the photoelectric conversion is a long-term goal that can be pursued via the fabrication of devices with ad-hoc architectures. In this framework, it is of utter importance to harvest and transform light irradiation into high electric potential inspecific area for driving functional dielectrics that respond to pure electric field. Here, a nano-fabrication technology has been devised featuring double self-alignment that is applied to construct zebra-like asymmetric heterojunction arrays. Such nanostructured composite, which covers a surface area of 5×4mm2 and contains 500 periodic repeating units, is capable of photo generating voltages as high as 140V on a flexible substrate. This approach represents a leap over the traditional functionalization process based on simply embedding materials into devices by demonstrating the disruptive potential of integrating oriented nanoscale device components into meta-material.

3DeepM: An Ad Hoc Architecture Based on Deep Learning Methods for Multispectral Image Classification

Current predefined architectures for deep learning are computationally very heavy and use tens of millions of parameters. Thus, computational costs may be prohibitive for many experimental or technological setups. We developed an ad hoc architecture for the classification of multispectral images using deep learning techniques. The architecture, called 3DeepM, is composed of 3D filter banks especially designed for the extraction of spatial-spectral features in multichannel images. The new architecture has been tested on a sample of 12210 multispectral images of seedless table grape varieties: Autumn Royal, Crimson Seedless, Itum4, Itum5 and Itum9. 3DeepM was able to classify 100% of the images and obtained the best overall results in terms of accuracy, number of classes, number of parameters and training time compared to similar work. In addition, this paper presents a flexible and reconfigurable computer vision system designed for the acquisition of multispectral images in the range of 400 nm to 1000 nm. The vision system enabled the creation of the first dataset consisting of 12210 37-channel multispectral images (12 VIS + 25 IR) of five seedless table grape varieties that have been used to validate the 3DeepM architecture. Compared to predefined classification architectures such as AlexNet, ResNet or ad hoc architectures with a very high number of parameters, 3DeepM shows the best classification performance despite using 130-fold fewer parameters than the architecture to which it was compared. 3DeepM can be used in a multitude of applications that use multispectral images, such as remote sensing or medical diagnosis. In addition, the small number of parameters of 3DeepM make it ideal for application in online classification systems aboard autonomous robots or unmanned vehicles.

Efficient Hardware Design of Iterative Stencil Loops

A large number of algorithms for multidimensional signals processing and scientific computation come in the form of iterative stencil loops (ISLs), whose data dependencies span across multiple iterations. Because of their complex inner structure, automatic hardware acceleration of such algorithms is traditionally considered as a difficult task. In this paper, we introduce an automatic design flow that identifies, in a wide family of bidimensional data processing algorithms, subportions that exhibit a kind of parallelism close to that of ISLs; these are mapped onto a space of highly optimized ad-hoc architectures, which is efficiently explored to identify the best implementations with respect to both area and throughput. Experimental results show that the proposed methodology generates circuits whose performance is comparable to that of manually optimized solutions, and orders of magnitude higher than those generated by commercial high-level synthesis tools.

Security in Ad Hoc Networks

Ad hoc networks are a wireless networking paradigm for mobile hosts. Unlike traditional mobile wireless networks, ad hoc networks do not rely on any fixed infrastructure. Instead, these networks are self-configurable and autonomous systems which are able to support movability and organize themselves arbitrarily. These unique characteristics of ad hoc networks pose a number of challenges for the implementation of security infrastructure in the wireless network system design. In this paper, we study the ad-hoc architecture thus understanding the vulnerabilities and security goals. Further, we discuss the various security attacks and explore approaches to secure the communication.

Fixed-point trigonometric functions on FPGAs

Three approaches for computing sines and cosines on FPGAs are studied in this paper, with a focus of highthroughput pipelined architecture, and state-of-the-art implementation techniques. The first approach is the classical CORDIC iteration, for which we suggest a reduced iteration technique and fine optimizations in datapath width and latency. The second is an ad-hoc architecture specifically designed around trigonometric identities. The third uses a generic table- and DSP-based polynomial approximator. These three architectures are implemented and compared in the FloPoCo framework.

Issues in the Usage of MANETs as Backup or Load-Balancing Transit Networks of the Internet

Internetworking the mobile ad-hoc networks (MANETs) with the Internet has been a hot issue for many years. However, most researches have been concentrated on the use of MANETs as the access networks for the Internet. This paper introduces another use of MANETs: backup or load-balancing transit networks for the Internet. Although the MANETs capacity is currently low compared with other backbone technologies, we argue that MANETs is a suitable alternative for backup or load-balancing transit networks of the Internet based on its self-organized infrastructure-less multi-hop architecture and its increasing capacity in the near future. To provide the backup or load-balancing transit services for the Internet, a scalable, stable, lowoverhead, QoS-support ad-hoc routing architecture with the address auto-configuration is required. Moreover, how an Internet gateway selects an external route via MANETs to another autonomous system (AS) also needs to be solved. In this paper, BGP-GCR+, a combination of the border gateway protocol (BGP), the gravitational cluster routing (GCR), and the passive/weak IPv6-based address stateless auto-configuration, is developed towards the standards to achieve the required functions. We have implemented of the routing component of our proposed BGP-GCR+ routing architecture [1] into ns-2 for the performance evaluation. Results show that the packet delivery ratio under high traffic load and mobility in BGPGCR+ is lower than that in AODV-Hello for small-scale (50-150 nodes), but slightly higher for the medium-scale (200 nodes) MANETs. For lower mobility and traffic load, the performance of both BGP-GCR+ and AODV-Hello are high (over 90%). Moreover, the routing overhead is much lower (almost one-half) in BGP-GCR+ compared with AODV in all test cases. The implementation of other components of BGP-GCR+ in ns-2 and the mapping of medium access control (MAC)-related metrics into the link stability of the cluster construction in BGP-GCR+ for quality of service (QoS) support, their performance evaluations are subject to our future researches.

Mobility, sensing, and security management in wireless ad hoc sensor systems

The development of various network architectures to connect wireless sensors for monitoring physical environments has emerged as an important new area of applications. The major issues occurring in these new types of systems are: (a) the continuity of environment monitoring, (b) the mobility management of sensors, (c) the ad-hoc communication scheme, and (d) the security of sensitive messages. This paper develops an ad-hoc architecture and communication and mobility schemes to provide an efficient monitoring system in an environment where targets have unpredicted motion. It also discusses a security solution. A numerical simulation is provided for the validation of our schemes.

WLAN standards and evolutions

Wireless Local Area Networks technologies have known an important technological and commercial development. Multiplicity of standards and variety of domains of use make necessary to compose with different technologies that can be seen either as concurrent or complementary. In this article, after positioning the different types of wireless networks (IEEE 802.11, HomeRF,HIPERLAN/2, Bluetooth) for mass market and professional applications destination, some generalities are briefly reminded such as centralised and ad-hoc architectures, regulatory constraints in the 2.45 and 5GHZ frequency bands used forWLAN, typical ranges, mobility and security features and limitations. Then the differentIEEE (802.11, 802.11a et 802.11b) andETSI (HIPERLAN/2) standards are described in details as well as their foreseen evolutions. It appears that 802.11 family of standards would take benefit of the currently existing products to evolve smoothly while integrating new features (broadband 802.11a physical layer, necessary radio features to meet European regulatory requirements, future introduction of Quality of Service schemes…). In the meantime,HIPERLAN/2 which has been specified as a complete system already supports most of those important features and is able to be adapted to various kinds of higher network layers. Lastly, it is shown that interworking schemes between 3G cellular systems andWLAN currently under investigations in 3GPP andETSI BRAN should permit in the future to easily operate wide area and multi-access technology based mobile networks.