Search

A continuing rise in DNN usage in distributed and embedded use cases has demanded more efficient hardware execution in the field. Low-precision GeMMs with optimized data formats have played a key role in more memory and computationally-efficient networks. Recently trending formats are block-scaled representations stemming from tight HW-SW co-optimization, that compress network size by sharing exponents per data block. Prior work mostly focuses on deploying such block-scaled GeMM operations on domain-specific accelerators for optimum efficiency at the cost of flexibility and ease of deployment. In this work, we exploit and optimize the deployment of block-scaled GeMMs on fully-programmable in-order vector processors using ARM SVE. We define a systematic methodology for performing design space exploration to optimally match the workload specifications with processor vector-lengths, different microkernels, block sizes and shapes. We introduce efficient intrinsics-based microkernels with effective loop unrollings, and data-transfer efficient fused requantization strategies to maximize kernel performance, while also ensuring several deployment configurations. We enable generalized block-scaled kernel deployments through tunable block sizes and shapes, which helps in accommodating different accuracy-speed trade-off requirements. Utilizing 2D activation blocks instead of conventional 1D blocks, the static and dynamic BS-INT8 configurations yielded on average 3.8x and 2.9x faster speedups over FP32 models respectively, at no accuracy loss for CNN classification tasks on CIFAR10/100 datasets. [ABSTRACT FROM AUTHOR]

Dr Marian Verhelst from KU Leuven in Belgium talks about the research behind her Letter 'Area scaling analysis of CMOS ADCs' on page 314. [ABSTRACT FROM AUTHOR]

In this paper, we present a simple analytic method that can be used to predict potential energy depletion in off-grid wireless backbone network nodes serving mobile users. The instantaneous energy depletion of the batteries of the network nodes is determined by random energy arrivals and departures and modeled as a G/G/1 queue. To evaluate the online energy depletion probability (EDP), an integral-free asymptotic approach is typically used by assuming that the prediction horizon approaches infinity. However, in many practical cases, the time required by user connections can be rather short. This indicates the need for proactive resource management decisions over finite horizons so that the transmission opportunities with limited energy and time horizons are not wasted. Using Hölder's inequality, we obtain a novel finite-horizon upper bound for the EDP, and the result is compared with the infinite-horizon method. The accuracy of the proposed bound, which is addressed both analytically and numerically, proves to be better for shorter prediction horizons. The finite- and infinite-horizon methods are then applied for an energy provisioning admission control (EP-AC) framework using off-grid backbone network nodes. The key observation of this paper is that the proposed finite-horizon prediction approach admits significantly more users to the network when the connection times are relatively short, while retaining an integral-free closed-form structure suitable for the online evaluation of the EDP. [ABSTRACT FROM AUTHOR]

Building efficient embedded deep learning systems requires a tight co-design between DNN algorithms, hardware, and algorithm-to-hardware mapping, a.k.a. dataflow. However, owing to the large joint design space, finding an optimal solution through physical implementation becomes infeasible. To tackle this problem, several design space exploration (DSE) frameworks have emerged recently, yet they either suffer from long runtimes or a limited exploration space. This article introduces ZigZag, a rapid DSE framework for DNN accelerator architecture and mapping. ZigZag extends the common DSE with uneven mapping opportunities and smart mapping search strategies. Uneven mapping decouples operands (W/I/O), memory hierarchy, and mappings (temporal/spatial), opening up a whole new space for DSE, and thus better design points are found by ZigZag compared to other SotAs. For this, ZigZag uses an enhanced nested-for-loop format as a uniform representation to integrate algorithm, accelerator, and algorithm-to-accelerator mapping. ZigZag consists of three key components: 1) an analytical energy-performance-area Hardware Cost Estimator, 2) two Mapping Search Engines that support spatial and temporal even/uneven mapping search, and 3) an Architecture Generator that auto-explores the wide memory hierarchy design space. Benchmarking experiments against published works, in-house accelerator, and existing DSE frameworks, together with three case studies, show the reliability and capability of ZigZag. Up to 64 percent more energy-efficient solutions are found compared to other SotAs, due to ZigZag's uneven mapping capabilities. [ABSTRACT FROM AUTHOR]

With the massive growth of wireless networks comes a bigger impact of collisions and interference, which has a negative effect on throughput and energy efficiency. To deal with this problem, we propose an in-band wireless collision and interference detection scheme based on full-duplex technology. To study its performance, we compare its throughput and energy efficiency with the performance of traditional half-duplex and symmetric in-band full-duplex transmissions. Our analysis considers a realistic protocol and overhead modeling, and a measurement-based self-interference model. Our results indicate that our proposed collision detection scheme can provide significant gains in terms of throughput and energy efficiency in large wireless networks. Moreover, when compared to half-duplex and symmetric full-duplex, our analysis shows that this scheme allows up to 45% more nodes in the network for the same energy consumption per bit. These results suggest that this could be an enabling technology towards efficient, dense wireless networks. [ABSTRACT FROM AUTHOR]

• Fusion rules for distributed adaptive node-specific signal estimation (DANSE) are not optimal when communication links are noisy in a wireless sensor network. • Fusion rules which take noisy links into account are developed, resulting in new algorithm named N-DANSE. • The strategy to prove convergence of N-DANSE is different from the strategy used for DANSE without noisy links, and includes the latter as a special case. • N-DANSE performs better than DANSE in scenarios with noisy links. • N-DANSE is consistently closer to the optimal performance than DANSE. We consider a distributed signal estimation problem in a wireless sensor network where each node aims to estimate a node-specific desired signal using all sensor signals available in the network. In this setting, the distributed adaptive node-specific signal estimation (DANSE) algorithm is able to learn optimal fusion rules with which the nodes fuse their sensor signals, as the fused signals are then transmitted between the nodes. Under the assumption of transmission without errors, DANSE achieves the performance of centralized estimation. However, noisy communication links introduce errors in these transmitted signals, e.g., due to quantization or communication errors. In this paper we show fusion rules which take additive noise in the transmitted signals into account at almost no increase in computational complexity, resulting in a new algorithm denoted as 'noisy-DANSE' (N-DANSE). As the convergence proof for DANSE cannot be straightforwardly generalized to the case with noisy links, we use a different strategy to prove convergence of N-DANSE, which also proves convergence of DANSE without noisy links as a special case. We validate the convergence of N-DANSE and compare its performance with the original DANSE through numerical simulations, which demonstrate the superiority of N-DANSE over the original DANSE in noisy links scenarios. [ABSTRACT FROM AUTHOR]

Knowledge of the performance of microbial electrolysis cells under a wide range of operating conditions is crucial to achieve high production efficiencies. Characterizing this performance in an experiment, however, is challenging due to either the long measurement times of steady-state procedures or the transient errors of dynamic procedures. Moreover, wide parallelization of the measurements is not feasible due to the high measurement equipment cost per channel. Hence, to speedup this characterization and to facilitate low-cost, yet widely parallel measurements, this paper presents a novel rapid polarization curve measurement procedure with a dynamic measurement resolution that runs on a custom six-channel potentiostat with a current-driven topology. As case study, the procedure is used to rapidly assess the impact of altering pH values on a microbial electrolysis cell that produces H2. A $\times 2$ – $\times 12$ speedup could be obtained in comparison with the state-of-the-art, depending on the characterization resolution (16–128 levels). On top of this speedup, measurements can be parallelized up to $6\times $ on the presented, affordable—42$-per-channel—potentiostat. [ABSTRACT FROM AUTHOR]

Telecommunication standards are documents that contain consolidated knowledge about communication systems and implementation best practices. They are created based on long consensus processes in order to meet practical constraints. This article describes how the DVB-S2 standard is used in the electrical engineering curricula at KU Leuven within a design and implementation course called EAGLE. The goal of the course is to teach third-year Bachelor students how to apply the abstract knowledge gathered in the theoretical courses (control theory, communication theory, software, and hardware) to the design of a complex and real engineering project. Based on their progress during the academic year, we illustrate how standards can be used to teach a broad skill set and act as an accelerator to abstract system implementation complexity. We give an overview of the objectives of the EAGLE project and the relevance of the DVB-S2 standard, the design and implementation task that the students need to accomplish, and how they finally learn to master the complexity of a real communication standard, VHDL coding, and system verification and integration. [ABSTRACT FROM PUBLISHER]

The interactions between three or more random variables are often nontrivial, poorly understood and, yet, are paramount for future advances in fields such as network information theory, neuroscience and genetics. In this work, we analyze these interactions as different modes of information sharing. Towards this end, and in contrast to most of the literature that focuses on analyzing the mutual information, we introduce an axiomatic framework for decomposing the joint entropy that characterizes the various ways in which random variables can share information. Our framework distinguishes between interdependencies where the information is shared redundantly and synergistic interdependencies where the sharing structure exists in the whole, but not between the parts. The key contribution of our approach is to focus on symmetric properties of this sharing, which do not depend on a specific point of view for differentiating roles between its components. We show that our axioms determine unique formulas for all of the terms of the proposed decomposition for systems of three variables in several cases of interest. Moreover, we show how these results can be applied to several network information theory problems, providing a more intuitive understanding of their fundamental limits. [ABSTRACT FROM AUTHOR]

This paper introduces an approach that combines machine learning and adaptive hardware to improve the efficiency of ultra-low-power sensor interfaces. Adaptive feature extraction circuits are assisted by hardware embedded training to dynamically activate only the most relevant features. This selection is done in a context- and power cost-aware manner, through modification of the C4.5 algorithm. As proof-of-principle, a Voice Activity Detector illustrates the context-dependent relevance of features, demonstrating average circuit power savings of 70%, without accuracy loss. The RECAS database developed for experimenting with this context- and dynamic resource-cost-aware training is presented and made open-source for the research community. [ABSTRACT FROM AUTHOR]

Testing radio frequency (RF) transceivers requires the measurement of a diverse set of specifications, requiring multiple testing setups. This complicates load board design, debug, and diagnosis, as well as results in long testing time. In this paper, we present a single setup testing solution for orthogonal frequency-division multiplexing systems RF front-ends based on a loop-around scheme. With this technique, it is possible to determine gain and phase mismatch, inphase–quadrature time skew, and dc offset. Linear gain and IIP3 decouple the transmitter parameters from the receiver parameters. Although loop-around has been used in many forms, the basic challenge is to determine what input conditions will lead to accurate measurement and what form of modeling will yield this accuracy. To this end, we develop test signal design and multistep extraction techniques. Experimental results indicate that IIP3 can be extracted with 0.6 dB maximum error while phase mismatch and gain mismatch can be extracted with 0.3^\circ and 0.6% maximum error. Our method is able to de-embed the characteristics of transmitter from those of receiver while it requires the analysis of only low-frequency digital baseband signals (I and Q branches) and eliminates the need for RF testers. [ABSTRACT FROM PUBLISHER]

The presence of microorganisms performing extracellular electron transfer has been established in many environments. Research to determine their role is moving slowly due to the high cost of potentiostats and the variance of data with small number of replicates. Here, we present a 128-channel potentiostat, connected to a 128 gold electrode array. Whereas the system is able to perform simultaneously 128 (bio)electrochemical measurements with an independent electrical signal input, the present manufacturing of the array limited the number of effective channels for this study to 77. We assessed the impact of 11 electrode potentials ranging from −0.45 V to +0.2 V vs. Ag/AgCl (7 replicates per potential) on the growth and electrochemical characteristics of anodic electroactive biofilms (EABs) formed by acetate-fed microbial communities. After 7 days of growth, maximum current was reached for electrodes poised at −0.3 V, closely followed by −0.25 V and −0.1 V to +0.1 V, a range well-fitting the midpoint potential of minerals naturally reduced by electroactive bacteria such as Geobacter Sulfurreducens. There was no significant difference in apparent midpoint potential of the EABs (−0.35 V), suggesting that the mechanism of heterogeneous electron transfer was not affected by the electrode potential. The EABs poised below current plateau potential (≤ −0.3 V) exhibited slower growth but higher charge transfer parameters. The high-throughput and high reproducibility provided by the array may have a major facilitating impact on the field of electromicrobiology. Key aspects to improve are data processing algorithms to deal with the vast amount of generated data, and manufacturing of the electrode array itself. • A 128-channel potentiostat for high-throughput microbial electrochemistry. • Accompanying 128 gold electrode array (77 functional electrodes). • Anodic electroactive biofilms simultaneously grown at 11 electrode potentials (n = 7). • Midpoint potentials and charge transport parameters assessed by cyclic voltammetry. • Most performant EABs grown just below anodic plateau (−0.3 V and −0.25 V vs. Ag/AgCl). [ABSTRACT FROM AUTHOR]

In this article, we present the simulation, fabrication, and characterization of a novel bilayer graphene field-effect transistor exhibiting electron mobility up to ∼1600 cm2 V−1 s−1, a room temperature Ion/Ioff ≈ 60, and the lowest total charge (∼1011 cm−2) reported to date. This is achieved by combined electrostatic and chemical doping of bilayer graphene, which enables one to switch off the device at zero top-gate voltage. Using density functional theory and atomistic simulations, we obtain physical insight into the impact of chemical and electrostatic doping on bandgap opening of bilayer graphene and the effect of metal contacts on the operation of the device. Our results represent a step forward in the use of bilayer graphene for high-performance logic devices in the beyond-complementary metal−oxide−semiconductor (CMOS) technology paradigm. [ABSTRACT FROM AUTHOR]