Search

The continuous improvement of fabrication techniques and high-quality semiconductor-superconductor interfaces allowed for unprecedented tunability of Josephson junction arrays (JJA), making them a promising candidate for analog quantum simulations of many-body phenomena. While most experimental proposals so far focused on quantum simulations of ensembles of two-level systems, the possibility of tuning the current-phase relation beyond the sinusoidal regime paves the way for studying statistical physics models with larger local Hilbert spaces. Here, we investigate a particular JJA architecture that can be mapped into a $\mathbb{Z}_3$ clock model. Through matrix-product-states simulations and bosonization analysis, we show that few experimentally accessible control parameters allow for the exploration of the rich phase diagrams of the associated low-energy field theories. Our results expand the horizon for analog quantum simulations with JJAs towards models that can not be efficiently captured with qubit architectures., Comment: 13+5 pages, 10+2 figures

With the rise of Embodied Foundation Models (EFMs), most notably Small Language Models (SLMs), adapting Transformers for edge applications has become a very active field of research. However, achieving end-to-end deployment of SLMs on microcontroller (MCU)-class chips without high-bandwidth off-chip main memory access is still an open challenge. In this paper, we demonstrate high-efficiency end-to-end SLM deployment on a multicore RISC-V (RV32) MCU augmented with ML instruction extensions and a hardware neural processing unit (NPU). To automate the exploration of the constrained, multi-dimensional memory vs. computation tradeoffs involved in aggressive SLM deployment on heterogeneous (multicore+NPU) resources, we introduce Deeploy, a novel Deep Neural Network (DNN) compiler, which generates highly-optimized C code requiring minimal runtime support. We demonstrate that Deeploy generates end-to-end code for executing SLMs, fully exploiting the RV32 cores' instruction extensions and the NPU: We achieve leading-edge energy and throughput of \SI{490}{\micro\joule \per Token}, at \SI{340}{Token \per \second} for an SLM trained on the TinyStories dataset, running for the first time on an MCU-class device without external memory., Comment: Accepted for publication at ESWEEK - CASES 2024

One of the challenges for Tiny Machine Learning (tinyML) is keeping up with the evolution of Machine Learning models from Convolutional Neural Networks to Transformers. We address this by leveraging a heterogeneous architectural template coupling RISC-V processors with hardwired accelerators supported by an automated deployment flow. We demonstrate an Attention-based model in a tinyML power envelope with an octa-core cluster coupled with an accelerator for quantized Attention. Our deployment flow enables an end-to-end 8-bit MobileBERT, achieving leading-edge energy efficiency and throughput of 2960 GOp/J and 154 GOp/s at 32.5 Inf/s consuming 52.0 mW (0.65 V, 22 nm FD-SOI technology)., Comment: Pre-print manuscript submitted for review to the IEEE Design and Test Special Issue on tinyML

The resource requirements of deep neural networks (DNNs) pose significant challenges to their deployment on edge devices. Common approaches to address this issue are pruning and mixed-precision quantization, which lead to latency and memory occupation improvements. These optimization techniques are usually applied independently. We propose a novel methodology to apply them jointly via a lightweight gradient-based search, and in a hardware-aware manner, greatly reducing the time required to generate Pareto-optimal DNNs in terms of accuracy versus cost (i.e., latency or memory). We test our approach on three edge-relevant benchmarks, namely CIFAR-10, Google Speech Commands, and Tiny ImageNet. When targeting the optimization of the memory footprint, we are able to achieve a size reduction of 47.50% and 69.54% at iso-accuracy with the baseline networks with all weights quantized at 8 and 2-bit, respectively. Our method surpasses a previous state-of-the-art approach with up to 56.17% size reduction at iso-accuracy. With respect to the sequential application of state-of-the-art pruning and mixed-precision optimizations, we obtain comparable or superior results, but with a significantly lowered training time. In addition, we show how well-tailored cost models can improve the cost versus accuracy trade-offs when targeting specific hardware for deployment.

This study presents a novel approach for EEG-based seizure detection leveraging a BERT-based model. The model, BENDR, undergoes a two-phase training process. Initially, it is pre-trained on the extensive Temple University Hospital EEG Corpus (TUEG), a 1.5 TB dataset comprising over 10,000 subjects, to extract common EEG data patterns. Subsequently, the model is fine-tuned on the CHB-MIT Scalp EEG Database, consisting of 664 EEG recordings from 24 pediatric patients, of which 198 contain seizure events. Key contributions include optimizing fine-tuning on the CHB-MIT dataset, where the impact of model architecture, pre-processing, and post-processing techniques are thoroughly examined to enhance sensitivity and reduce false positives per hour (FP/h). We also explored custom training strategies to ascertain the most effective setup. The model undergoes a novel second pre-training phase before subject-specific fine-tuning, enhancing its generalization capabilities. The optimized model demonstrates substantial performance enhancements, achieving as low as 0.23 FP/h, 2.5$\times$ lower than the baseline model, with a lower but still acceptable sensitivity rate, showcasing the effectiveness of applying a BERT-based approach on EEG-based seizure detection., Comment: 4 pages, 2 tables, 2 figures

Depthwise separable convolutions are a fundamental component in efficient Deep Neural Networks, as they reduce the number of parameters and operations compared to traditional convolutions while maintaining comparable accuracy. However, their low data reuse opportunities make deploying them notoriously difficult. In this work, we perform an extensive exploration of alternatives to fuse the depthwise and pointwise kernels that constitute the separable convolutional block. Our approach aims to minimize time-consuming memory transfers by combining different data layouts. When targeting a commercial ultra-low-power device with a three-level memory hierarchy, the GreenWaves GAP8 SoC, we reduce the latency of end-to-end network execution by up to 11.40%. Furthermore, our kernels reduce activation data movements between L2 and L1 memories by up to 52.97%., Comment: Accepted at the XXIV International Conference on Embedded Computer Systems: Architectures, Modeling and Simulation (SAMOS2024), June 29 - July 4, 2024

Optimal deployment of deep neural networks (DNNs) on state-of-the-art Systems-on-Chips (SoCs) is crucial for tiny machine learning (TinyML) at the edge. The complexity of these SoCs makes deployment non-trivial, as they typically contain multiple heterogeneous compute cores with limited, programmer-managed memory to optimize latency and energy efficiency. We propose HTVM - a compiler that merges TVM with DORY to maximize the utilization of heterogeneous accelerators and minimize data movements. HTVM allows deploying the MLPerf(TM) Tiny suite on DIANA, an SoC with a RISC-V CPU, and digital and analog compute-in-memory AI accelerators, at 120x improved performance over plain TVM deployment., Comment: Presented at DAC2023. Open-source code is available at https://github.com/KULeuven-MICAS/htvm

Transformer-based foundation models have become crucial for various domains, most notably natural language processing (NLP) or computer vision (CV). These models are predominantly deployed on high-performance GPUs or hardwired accelerators with highly customized, proprietary instruction sets. Until now, limited attention has been given to RISC-V-based general-purpose platforms. In our work, we present the first end-to-end inference results of transformer models on an open-source many-tiny-core RISC-V platform implementing distributed Softmax primitives and leveraging ISA extensions for SIMD floating-point operand streaming and instruction repetition, as well as specialized DMA engines to minimize costly main memory accesses and to tolerate their latency. We focus on two foundational transformer topologies, encoder-only and decoder-only models. For encoder-only models, we demonstrate a speedup of up to 12.8x between the most optimized implementation and the baseline version. We reach over 79% FPU utilization and 294 GFLOPS/W, outperforming State-of-the-Art (SoA) accelerators by more than 2x utilizing the HW platform while achieving comparable throughput per computational unit. For decoder-only topologies, we achieve 16.1x speedup in the Non-Autoregressive (NAR) mode and up to 35.6x speedup in the Autoregressive (AR) mode compared to the baseline implementation. Compared to the best SoA dedicated accelerator, we achieve 2.04x higher FPU utilization., Comment: 14 pages, 10 figures, 4 tables, IEEE Transactions on Circuits and Systems for Artificial Intelligence

We investigate the dynamics of dilute systems composed of nucleons and light clusters within a linear response approach, taking into account the in-medium Mott effects on cluster appearance, through a density-dependent momentum cut-off. We find that spinodal instabilities and associated growth rates are severely affected by the presence of light clusters and, in particular, by the treatment of in-medium effects, foreshadowing intriguing consequences for fragment formation in heavy-ion collisions and in the broader astrophysical context., Comment: 6 pages, 4 figures

Transformer networks are rapidly becoming SotA in many fields, such as NLP and CV. Similarly to CNN, there is a strong push for deploying Transformer models at the extreme edge, ultimately fitting the tiny power budget and memory footprint of MCUs. However, the early approaches in this direction are mostly ad-hoc, platform, and model-specific. This work aims to enable and optimize the flexible, multi-platform deployment of encoder Tiny Transformers on commercial MCUs. We propose a complete framework to perform end-to-end deployment of Transformer models onto single and multi-core MCUs. Our framework provides an optimized library of kernels to maximize data reuse and avoid unnecessary data marshaling operations into the crucial attention block. A novel MHSA inference schedule, named Fused-Weight Self-Attention, is introduced, fusing the linear projection weights offline to further reduce the number of operations and parameters. Furthermore, to mitigate the memory peak reached by the computation of the attention map, we present a Depth-First Tiling scheme for MHSA. We evaluate our framework on three different MCU classes exploiting ARM and RISC-V ISA, namely the STM32H7, the STM32L4, and GAP9 (RV32IMC-XpulpV2). We reach an average of 4.79x and 2.0x lower latency compared to SotA libraries CMSIS-NN (ARM) and PULP-NN (RISC-V), respectively. Moreover, we show that our MHSA depth-first tiling scheme reduces the memory peak by up to 6.19x, while the fused-weight attention can reduce the runtime by 1.53x, and number of parameters by 25%. We report significant improvements across several Tiny Transformers: for instance, when executing a transformer block for the task of radar-based hand-gesture recognition on GAP9, we achieve a latency of 0.14ms and energy consumption of 4.92 micro-joules, 2.32x lower than the SotA PULP-NN library on the same platform., Comment: Pre-print manuscript submitted for review to the IEEE Transactions on Computers

Structural Health Monitoring (SHM) is a critical task for ensuring the safety and reliability of civil infrastructures, typically realized on bridges and viaducts by means of vibration monitoring. In this paper, we propose for the first time the use of Transformer neural networks, with a Masked Auto-Encoder architecture, as Foundation Models for SHM. We demonstrate the ability of these models to learn generalizable representations from multiple large datasets through self-supervised pre-training, which, coupled with task-specific fine-tuning, allows them to outperform state-of-the-art traditional methods on diverse tasks, including Anomaly Detection (AD) and Traffic Load Estimation (TLE). We then extensively explore model size versus accuracy trade-offs and experiment with Knowledge Distillation (KD) to improve the performance of smaller Transformers, enabling their embedding directly into the SHM edge nodes. We showcase the effectiveness of our foundation models using data from three operational viaducts. For AD, we achieve a near-perfect 99.9% accuracy with a monitoring time span of just 15 windows. In contrast, a state-of-the-art method based on Principal Component Analysis (PCA) obtains its first good result (95.03% accuracy) only considering 120 windows. On two different TLE tasks, our models obtain state-of-the-art performance on multiple evaluation metrics (R$^2$ score, MAE% and MSE%). On the first benchmark, we achieve an R$^2$ score of 0.97 and 0.85 for light and heavy vehicle traffic, respectively, while the best previous approach stops at 0.91 and 0.84. On the second one, we achieve an R$^2$ score of 0.54 versus the 0.10 of the best existing method., Comment: 16 pages, 4 tables, 9 figures

RISC-V cores have gained a lot of popularity over the last few years. However, being quite a recent and novel technology, there is still a gap in the availability of comprehensive simulation frameworks for RISC-V that cover both the functional and extra-functional aspects. This gap hinders progress in the field, as fast yet accurate system-level simulation is crucial for Design Space Exploration (DSE). This work presents an open-source framework designed to tackle this challenge, integrating functional RISC-V simulation (achieved with GVSoC) with SystemC-AMS (used to model extra-functional aspects, in detail power storage and distribution). The combination of GVSoC and SystemC-AMS in a single simulation framework allows to perform a DSE that is dependent on the mutual impact between functional and extra-functional aspects. In our experiments, we validate the framework's effectiveness by creating a virtual prototype of a compact, battery-powered embedded system., Comment: 4 pages, 4 figures, to be published in Computing Frontiers Workshops and Special Sessions (CF '24 Companion), May 7--9, 2024, Ischia, Italy

Recently, efficiently deploying deep learning solutions on the edge has received increasing attention. New platforms are emerging to support the increasing demand for flexibility and high performance. In this work, we explore the efficient mapping of convolutional layers on an open-hardware, low-power Coarse-Grain Reconfigurable Array (CGRA), namely OpenEdgeCGRA. We explore both direct implementations of convolution and solutions that transform it into a matrix multiplication through an Im2col transformation, and experiment with various tensor parallelism axes. We show that for this hardware target, direct convolution, coupled with weight parallelism reaches the best latency and energy efficiency, outperforming a CPU implementation by 3.4x and 9.9x in terms of energy and latency, respectively.

Miniaturized autonomous unmanned aerial vehicles (UAVs) are gaining popularity due to their small size, enabling new tasks such as indoor navigation or people monitoring. Nonetheless, their size and simple electronics pose severe challenges in implementing advanced onboard intelligence. This work proposes a new automatic optimization pipeline for visual pose estimation tasks using Deep Neural Networks (DNNs). The pipeline leverages two different Neural Architecture Search (NAS) algorithms to pursue a vast complexity-driven exploration in the DNNs' architectural space. The obtained networks are then deployed on an off-the-shelf nano-drone equipped with a parallel ultra-low power System-on-Chip leveraging a set of novel software kernels for the efficient fused execution of critical DNN layer sequences. Our results improve the state-of-the-art reducing inference latency by up to 3.22x at iso-error., Comment: This paper has been accepted for publication in the ERF 2024 conference

Low-resolution infrared (IR) array sensors enable people counting applications such as monitoring the occupancy of spaces and people flows while preserving privacy and minimizing energy consumption. Deep Neural Networks (DNNs) have been shown to be well-suited to process these sensor data in an accurate and efficient manner. Nevertheless, the space of DNNs' architectures is huge and its manual exploration is burdensome and often leads to sub-optimal solutions. To overcome this problem, in this work, we propose a highly automated full-stack optimization flow for DNNs that goes from neural architecture search, mixed-precision quantization, and post-processing, down to the realization of a new smart sensor prototype, including a Microcontroller with a customized instruction set. Integrating these cross-layer optimizations, we obtain a large set of Pareto-optimal solutions in the 3D-space of energy, memory, and accuracy. Deploying such solutions on our hardware platform, we improve the state-of-the-art achieving up to 4.2x model size reduction, 23.8x code size reduction, and 15.38x energy reduction at iso-accuracy., Comment: This paper has been accepted for publication in the DATE 2024 conference IEEE

Sub-10cm diameter nano-drones are gaining momentum thanks to their applicability in scenarios prevented to bigger flying drones, such as in narrow environments and close to humans. However, their tiny form factor also brings their major drawback: ultra-constrained memory and processors for the onboard execution of their perception pipelines. Therefore, lightweight deep learning-based approaches are becoming increasingly popular, stressing how computational efficiency and energy-saving are paramount as they can make the difference between a fully working closed-loop system and a failing one. In this work, to maximize the exploitation of the ultra-limited resources aboard nano-drones, we present a novel adaptive deep learning-based mechanism for the efficient execution of a vision-based human pose estimation task. We leverage two State-of-the-Art (SoA) convolutional neural networks (CNNs) with different regression performance vs. computational costs trade-offs. By combining these CNNs with three novel adaptation strategies based on the output's temporal consistency and on auxiliary tasks to swap the CNN being executed proactively, we present six different systems. On a real-world dataset and the actual nano-drone hardware, our best-performing system, compared to executing only the bigger and most accurate SoA model, shows 28% latency reduction while keeping the same mean absolute error (MAE), 3% MAE reduction while being iso-latency, and the absolute peak performance, i.e., 6% better than SoA model., Comment: Accepted for publication in the 2024 Design, Automation and Test in Europe (DATE) conference

Modern hybrid superconductor-semiconductor Josephson junction arrays are a promising platform for analog quantum simulations. Their controllable and non-sinusoidal energy/phase relation opens the path to implement nontrivial interactions and study the emergence of exotic quantum phase transitions. Here, we propose the analysis of an array of hybrid Josephson junctions defining a 2-leg ladder geometry for the quantum simulation of the tricritical Ising phase transition. This transition provides the paradigmatic example of minimal conformal models beyond Ising criticality and its excitations are intimately related with Fibonacci non-Abelian anyons and topological order in two dimensions. We study this superconducting system and its thermodynamic phases based on bosonization and matrix-product-states techniques. Its effective continuous description in terms of a three-frequency sine-Gordon quantum field theory suggests the presence of the targeted tricritical point and the numerical simulations confirm this picture. Our results indicate which experimental observables can be adopted in realistic devices to probe the physics and the phase transitions of the model. Additionally, our proposal provides a useful one-dimensional building block to design exotic topological order in two-dimensional scalable Josephson junction arrays., Comment: 6 pages, 3 figures and Supplemental material

The rapid proliferation of computing domains relying on Internet of Things (IoT) devices has created a pressing need for efficient and accurate deep-learning (DL) models that can run on low-power devices. However, traditional DL models tend to be too complex and computationally intensive for typical IoT end-nodes. To address this challenge, Neural Architecture Search (NAS) has emerged as a popular design automation technique for co-optimizing the accuracy and complexity of deep neural networks. Nevertheless, existing NAS techniques require many iterations to produce a network that adheres to specific hardware constraints, such as the maximum memory available on the hardware or the maximum latency allowed by the target application. In this work, we propose a novel approach to incorporate multiple constraints into so-called Differentiable NAS optimization methods, which allows the generation, in a single shot, of a model that respects user-defined constraints on both memory and latency in a time comparable to a single standard training. The proposed approach is evaluated on five IoT-relevant benchmarks, including the MLPerf Tiny suite and Tiny ImageNet, demonstrating that, with a single search, it is possible to reduce memory and latency by 87.4% and 54.2%, respectively (as defined by our targets), while ensuring non-inferior accuracy on state-of-the-art hand-tuned deep neural networks for TinyML., Comment: Accepted for publication at the IEEE Transactions on Emerging Topics in Computing

Recent advancements in fabrication techniques have enabled unprecedented clean interfaces and gate tunability in semiconductor-superconductor heterostructures. Inspired by these developments, we propose protocols to realize Thouless quantum pumping in electrically tunable Josephson junction arrays. We analyze, in particular, the implementation of the Rice-Mele and the Harper-Hofstadter pumping schemes, whose realization would validate these systems as flexible platforms for quantum simulations. We investigate numerically the long-time behavior of chains of controllable superconducting islands in the Coulomb-blockaded regime. Our findings provide new insights into the dynamics of periodically driven interacting systems and highlight the robustness of Thouless pumping with respect to boundary effects typical of superconducting circuits., Comment: 26 pages, 12 figures

Accurate yet efficient Deep Neural Networks (DNNs) are in high demand, especially for applications that require their execution on constrained edge devices. Finding such DNNs in a reasonable time for new applications requires automated optimization pipelines since the huge space of hyper-parameter combinations is impossible to explore extensively by hand. In this work, we propose PLiNIO, an open-source library implementing a comprehensive set of state-of-the-art DNN design automation techniques, all based on lightweight gradient-based optimization, under a unified and user-friendly interface. With experiments on several edge-relevant tasks, we show that combining the various optimizations available in PLiNIO leads to rich sets of solutions that Pareto-dominate the considered baselines in terms of accuracy vs model size. Noteworthy, PLiNIO achieves up to 94.34% memory reduction for a <1% accuracy drop compared to a baseline architecture., Comment: Accepted at the 2023 Forum on Specification & Design Languages (FDL)

The topological Kondo effect is a genuine manifestation of the nonlocality of Majorana modes. We investigate its out-of-equilibrium signatures in a model with a Cooper-pair box hosting four of these topological modes, each connected to a metallic lead. Through an advanced matrix-product-state approach tailored to study the dynamics of superconductors, we simulate the relaxation of the Majorana magnetization, which allows us to determine the related Kondo temperature, and we analyze the onset of electric transport after a quantum quench of a lead voltage. Our results apply to Majorana Cooper-pair boxes fabricated in double nanowire devices and provide nonperturbative evidence of the crossover from weak-coupling states to the strongly correlated topological Kondo regime. The latter dominates at the superconductor charge degeneracy points and displays the expected universal fractional zero-bias conductance., Comment: 5 pages, 5 figures, Supplemental Material (8 pages, 8 figures)

The emerging trend of deploying complex algorithms, such as Deep Neural Networks (DNNs), increasingly poses strict memory and energy efficiency requirements on Internet-of-Things (IoT) end-nodes. Mixed-precision quantization has been proposed as a technique to minimize a DNN's memory footprint and maximize its execution efficiency, with negligible end-to-end precision degradation. In this work, we present a novel hardware and software stack for energy-efficient inference of mixed-precision Quantized Neural Networks (QNNs). We introduce Flex-V, a processor based on the RISC-V Instruction Set Architecture (ISA) that features fused Mac&Load mixed-precision dot product instructions; to avoid the exponential growth of the encoding space due to mixed-precision variants, we encode formats into the Control-Status Registers (CSRs). Flex-V core is integrated into a tightly-coupled cluster of eight processors; in addition, we provide a full framework for the end-to-end deployment of DNNs including a compiler, optimized libraries, and a memory-aware deployment flow. Our results show up to 91.5 MAC/cycle and 3.26 TOPS/W on the cluster, implemented in a commercial 22nm FDX technology, with up to 8.5x speed-up, and an area overhead of only 5.6% with respect to the baseline. To demonstrate the capabilities of the architecture, we benchmark it with end-to-end real-life QNNs, improving performance by 2x - 2.5x with respect to existing solutions using fully flexible programmable processors.

With the increasing popularity of Internet of Things (IoT) devices, there is a growing need for energy-efficient Machine Learning (ML) models that can run on constrained edge nodes. Decision tree ensembles, such as Random Forests (RFs) and Gradient Boosting (GBTs), are particularly suited for this task, given their relatively low complexity compared to other alternatives. However, their inference time and energy costs are still significant for edge hardware. Given that said costs grow linearly with the ensemble size, this paper proposes the use of dynamic ensembles, that adjust the number of executed trees based both on a latency/energy target and on the complexity of the processed input, to trade-off computational cost and accuracy. We focus on deploying these algorithms on multi-core low-power IoT devices, designing a tool that automatically converts a Python ensemble into optimized C code, and exploring several optimizations that account for the available parallelism and memory hierarchy. We extensively benchmark both static and dynamic RFs and GBTs on three state-of-the-art IoT-relevant datasets, using an 8-core ultra-lowpower System-on-Chip (SoC), GAP8, as the target platform. Thanks to the proposed early-stopping mechanisms, we achieve an energy reduction of up to 37.9% with respect to static GBTs (8.82 uJ vs 14.20 uJ per inference) and 41.7% with respect to static RFs (2.86 uJ vs 4.90 uJ per inference), without losing accuracy compared to the static model., Comment: This article has been accepted for publication in IEEE Internet of Things Journal

Modern smartwatches often include photoplethysmographic (PPG) sensors to measure heartbeats or blood pressure through complex algorithms that fuse PPG data with other signals. In this work, we propose a collaborative inference approach that uses both a smartwatch and a connected smartphone to maximize the performance of heart rate (HR) tracking while also maximizing the smartwatch's battery life. In particular, we first analyze the trade-offs between running on-device HR tracking or offloading the work to the mobile. Then, thanks to an additional step to evaluate the difficulty of the upcoming HR prediction, we demonstrate that we can smartly manage the workload between smartwatch and smartphone, maintaining a low mean absolute error (MAE) while reducing energy consumption. We benchmark our approach on a custom smartwatch prototype, including the STM32WB55 MCU and Bluetooth Low-Energy (BLE) communication, and a Raspberry Pi3 as a proxy for the smartphone. With our Collaborative Heart Rate Inference System (CHRIS), we obtain a set of Pareto-optimal configurations demonstrating the same MAE as State-of-Art (SoA) algorithms while consuming less energy. For instance, we can achieve approximately the same MAE of TimePPG-Small (5.54 BPM MAE vs. 5.60 BPM MAE) while reducing the energy by 2.03x, with a configuration that offloads 80\% of the predictions to the phone. Furthermore, accepting a performance degradation to 7.16 BPM of MAE, we can achieve an energy consumption of 179 uJ per prediction, 3.03x less than running TimePPG-Small on the smartwatch, and 1.82x less than streaming all the input data to the phone., Comment: Published at 2023 Design, Automation \& Test in Europe Conference \& Exhibition (DATE)

The need to execute Deep Neural Networks (DNNs) at low latency and low power at the edge has spurred the development of new heterogeneous Systems-on-Chips (SoCs) encapsulating a diverse set of hardware accelerators. How to optimally map a DNN onto such multi-accelerator systems is an open problem. We propose ODiMO, a hardware-aware tool that performs a fine-grain mapping across different accelerators on-chip, splitting individual layers and executing them in parallel, to reduce inference energy consumption or latency, while taking into account each accelerator's quantization precision to maintain accuracy. Pareto-optimal networks in the accuracy vs. energy or latency space are pursued for three popular dataset/DNN pairs, and deployed on the DIANA heterogeneous ultra-low power edge AI SoC. We show that ODiMO reduces energy/latency by up to 33%/31% with limited accuracy drop (-0.53%/-0.32%) compared to manual heuristic mappings., Comment: Accepted at 2023 ACM/IEEE International Symposium on Low Power Electronics and Design (ISLPED)

Advances in hybrid fractional quantum Hall (FQH)-superconductor platforms pave the way for realisation of parafermionic modes. We analyse signatures of these non-abelian anyons in transport measurements across devices with $\mathbb{Z}_6$ parafermions (PFs) coupled to an external electrode. Simulating the dynamics of these open systems by a stochastic quantum jump method, we show that a current readout over sufficiently long times constitutes a projective measurement of the fractional charge shared by two PFs. Interaction of these topological modes with the FQH environment, however, may cause poisoning events affecting this degree of freedom which we model by jump operators that describe incoherent coupling of PFs with FQH edge modes. We analyse how this gives rise to a characteristic three-level telegraph noise in the current, constituting a very strong signature of PFs. We discuss also other forms of poisoning and noise caused by interaction with fractional quasiparticles in the bulk of the Hall system. We conclude our work with an analysis of four-PF devices, in particular on how the PF fusion algebra can be observed in electrical transport experiments.

The Anderson impurity model is a paradigmatic example in the study of strongly correlated quantum systems and describes an interacting quantum dot coupled to electronic leads. In this work, we characterize the emergence of the Kondo effect by investigating the model dynamics following a quantum quench based on matrix product state simulations. The relaxation of the impurity magnetization allows for the estimate of the predicted universal scaling of the Kondo temperature as a function of the impurity-lead hybridization and quantum dot repulsion. Additionally, our simulations permit us to evaluate the current in the nonequilibrium quasi-steady state appearing after the quench. Through their values, we examine the dependence of the conductance on the voltage bias $V_b$ and on the impurity chemical potential $V_g$, which displays a zero-bias Kondo peak. Our results are relevant for transport measurements in Coulomb blockaded devices, and, in particular, in quantum dots induced in nanowires., Comment: 9 pages 5 figures

Miniaturized autonomous unmanned aerial vehicles (UAVs) are an emerging and trending topic. With their form factor as big as the palm of one hand, they can reach spots otherwise inaccessible to bigger robots and safely operate in human surroundings. The simple electronics aboard such robots (sub-100mW) make them particularly cheap and attractive but pose significant challenges in enabling onboard sophisticated intelligence. In this work, we leverage a novel neural architecture search (NAS) technique to automatically identify several Pareto-optimal convolutional neural networks (CNNs) for a visual pose estimation task. Our work demonstrates how real-life and field-tested robotics applications can concretely leverage NAS technologies to automatically and efficiently optimize CNNs for the specific hardware constraints of small UAVs. We deploy several NAS-optimized CNNs and run them in closed-loop aboard a 27-g Crazyflie nano-UAV equipped with a parallel ultra-low power System-on-Chip. Our results improve the State-of-the-Art by reducing the in-field control error of 32% while achieving a real-time onboard inference-rate of ~10Hz@10mW and ~50Hz@90mW., Comment: \c{opyright} 2023 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works

Neural Architecture Search (NAS) is quickly becoming the go-to approach to optimize the structure of Deep Learning (DL) models for complex tasks such as Image Classification or Object Detection. However, many other relevant applications of DL, especially at the edge, are based on time-series processing and require models with unique features, for which NAS is less explored. This work focuses in particular on Temporal Convolutional Networks (TCNs), a convolutional model for time-series processing that has recently emerged as a promising alternative to more complex recurrent architectures. We propose the first NAS tool that explicitly targets the optimization of the most peculiar architectural parameters of TCNs, namely dilation, receptive-field and number of features in each layer. The proposed approach searches for networks that offer good trade-offs between accuracy and number of parameters/operations, enabling an efficient deployment on embedded platforms. We test the proposed NAS on four real-world, edge-relevant tasks, involving audio and bio-signals. Results show that, starting from a single seed network, our method is capable of obtaining a rich collection of Pareto optimal architectures, among which we obtain models with the same accuracy as the seed, and 15.9-152x fewer parameters. Compared to three state-of-the-art NAS tools, ProxylessNAS, MorphNet and FBNetV2, our method explores a larger search space for TCNs (up to 10^12x) and obtains superior solutions, while requiring low GPU memory and search time. We deploy our NAS outputs on two distinct edge devices, the multicore GreenWaves Technology GAP8 IoT processor and the single-core STMicroelectronics STM32H7 microcontroller. With respect to the state-of-the-art hand-tuned models, we reduce latency and energy of up to 5.5x and 3.8x on the two targets respectively, without any accuracy loss., Comment: Accepted for publication at the IEEE Transactions on Computers

Nowadays, Hearth Rate (HR) monitoring is a key feature of almost all wrist-worn devices exploiting photoplethysmography (PPG) sensors. However, arm movements affect the performance of PPG-based HR tracking. This issue is usually addressed by fusing the PPG signal with data produced by inertial measurement units. Thus, deep learning algorithms have been proposed, but they are considered too complex to deploy on wearable devices and lack the explainability of results. In this work, we present a new deep learning model, PULSE, which exploits temporal convolutions and multi-head cross-attention to improve sensor fusion's effectiveness and achieve a step towards explainability. We evaluate the performance of PULSE on three publicly available datasets, reducing the mean absolute error by 7.56% on the most extensive available dataset, PPG-DaLiA. Finally, we demonstrate the explainability of PULSE and the benefits of applying attention modules to PPG and motion data., Comment: To be submitted to IEEE ICASSP 2023

Reading out the parity degree of freedom of Majorana bound states is key to demonstrating their nonabelian exchange properties. Here, we present a low-energy model describing localized edge states in a two-arm device. We study parity-to-charge conversion based on coupling the superconductor bound states to a quantum dot whose charge is read out by a sensor. The dynamics of the system, including the readout device, is analyzed in full using a quantum-jump approach. We show how the resulting signal and noise ratio differentiates between local Majorana and Andreev bound states., Comment: Main paper: 7 pages, 3 figures. Supplementary material: 10 pages, 7 figures

"Strange" correlators provide a tool to detect topological phases arising in many-body models by computing the matrix elements of suitably defined two-point correlations between the states under investigation and trivial reference states. Their effectiveness depends on the choice of the adopted operators. In this paper we give a systematic procedure for this choice, discussing the advantages of choosing operators using the bulk-boundary correspondence of the systems under scrutiny. Via the scaling exponents, we directly relate the algebraic decay of the strange correlators with the scaling dimensions of gapless edge modes operators. We begin our analysis with lattice models hosting symmetry-protected topological phases and we analyze the sums of the strange correlators, pointing out that integrating their moduli substantially reduces cancellations and finite-size effects. We also analyze instances of systems hosting intrinsic topological order, as well as strange correlators between states with different nontrivial topologies. Our results for both translational and non-translational invariant cases, and in presence of on-site disorder and long-range couplings, extend the validity of the strange correlators approach for the diagnosis of topological phases of matter, and indicate a general procedure for their optimal choice., Comment: 28 pages, 10 figures

Subgap states are responsible for the low-bias transport features of hybrid superconducting--semiconducting devices. Here, we analyze the local and nonlocal differential conductance of Coulomb-blockaded multiterminal superconducting islands that host subgap states with different spatial structures. The emerging patterns of their transport spectroscopy are used to characterize the possible topological nature of these devices and offer the possibility of controlling their transport properties. We develop a next-to-leading order master equation to describe the multiterminal transport in superconductors with both strong Coulomb interactions and multiple subgap states, coupled with metallic leads. We show that the nonlocal differential conductance characterizes the spatial extension of the subgap states and signals the presence of degenerate bound states with a finite support on different parts of the device. Additionally, it displays sharp sign changes as a function of the induced charge of the superconductor, signaling energy crossings among its lowest excited states., Comment: 14 pages, 7 figures

Human Activity Recognition (HAR) based on inertial data is an increasingly diffused task on embedded devices, from smartphones to ultra low-power sensors. Due to the high computational complexity of deep learning models, most embedded HAR systems are based on simple and not-so-accurate classic machine learning algorithms. This work bridges the gap between on-device HAR and deep learning, proposing a set of efficient one-dimensional Convolutional Neural Networks (CNNs) deployable on general purpose microcontrollers (MCUs). Our CNNs are obtained combining hyper-parameters optimization with sub-byte and mixed-precision quantization, to find good trade-offs between classification results and memory occupation. Moreover, we also leverage adaptive inference as an orthogonal optimization to tune the inference complexity at runtime based on the processed input, hence producing a more flexible HAR system. With experiments on four datasets, and targeting an ultra-low-power RISC-V MCU, we show that (i) We are able to obtain a rich set of Pareto-optimal CNNs for HAR, spanning more than 1 order of magnitude in terms of memory, latency and energy consumption; (ii) Thanks to adaptive inference, we can derive >20 runtime operating modes starting from a single CNN, differing by up to 10% in classification scores and by more than 3x in inference complexity, with a limited memory overhead; (iii) on three of the four benchmarks, we outperform all previous deep learning methods, reducing the memory occupation by more than 100x. The few methods that obtain better performance (both shallow and deep) are not compatible with MCU deployment. (iv) All our CNNs are compatible with real-time on-device HAR with an inference latency <16ms. Their memory occupation varies in 0.05-23.17 kB, and their energy consumption in 0.005 and 61.59 uJ, allowing years of continuous operation on a small battery supply.

Superconducting devices subject to strong charging energy interactions and Coulomb blockade are one of the key elements for the development of nanoelectronics and constitute common building blocks of quantum computation platforms and topological superconducting setups. The study of their transport properties is non-trivial and some of their non-perturbative aspects are hard to capture with the most ordinary techniques. Here we present a matrix product state approach to simulate the real-time dynamics of these systems. We propose a study of their transport based on the analysis of the currents after quantum quenches connecting such devices with external leads. Our method is based on the combination of a Wilson chain construction for the leads and a mean-field BCS description for the superconducting scatterers. In particular, we employ a quasiparticle energy eigenbasis which greatly reduces their entanglement growth and we introduce an auxiliary degree of freedom to encode the device total charge. This approach allows us to treat non-perturbatively both their charging energy and coupling with external electrodes. We show that our construction is able to describe the Coulomb diamond structure of a superconducting dot with subgap states, including its sequential tunneling and cotunneling features. We also study the conductance zero-bias peaks caused by Majorana modes in a blockaded Kitaev chain, and compare our results with common Breit-Wigner predictions., Comment: 21 pages, 10 figures

Purpose: We want to analyze transitions to low-lying excited states of the residual and ejectile nuclei in the 76Se(18O, 17O) 77Se one-neutron stripping reaction at 275-MeV incident energy and determine the role of single-particle and core excitation in the description of the measured cross sections. In addition, we explore the sensitivity of the calculated cross section to different nuclear structure models. Methods: The excitation energy spectrum and the differential cross-section angular distributions are measured using the MAGNEX large acceptance magnetic spectrometer for the detection of the ejectiles and the missing mass technique for the reconstruction of the reaction kinematics. The data are compared with calculations based on distorted-wave Born approximation, coupled-channels Born approximation, and coupled reaction channels adopting spectroscopic amplitudes for the projectile and target overlaps derived by large-scale shell-model calculations and interacting boson-fermion model. Results: Peaks in the energy spectra corresponding to groups of unresolved transitions to 77Se and 17O are identified. The experimental cross sections are extracted and compared to theoretical calculations. A remarkable agreement is found, without using any scaling factors, demonstrating that the adopted models for nuclear structure and reaction take into account the relevant aspects of the studied processes. The main transitions which contribute to the cross section of each peak are identified., Comment: 15 pages, 4 figures

Quantization is widely employed in both cloud and edge systems to reduce the memory occupation, latency, and energy consumption of deep neural networks. In particular, mixed-precision quantization, i.e., the use of different bit-widths for different portions of the network, has been shown to provide excellent efficiency gains with limited accuracy drops, especially with optimized bit-width assignments determined by automated Neural Architecture Search (NAS) tools. State-of-the-art mixed-precision works layer-wise, i.e., it uses different bit-widths for the weights and activations tensors of each network layer. In this work, we widen the search space, proposing a novel NAS that selects the bit-width of each weight tensor channel independently. This gives the tool the additional flexibility of assigning a higher precision only to the weights associated with the most informative features. Testing on the MLPerf Tiny benchmark suite, we obtain a rich collection of Pareto-optimal models in the accuracy vs model size and accuracy vs energy spaces. When deployed on the MPIC RISC-V edge processor, our networks reduce the memory and energy for inference by up to 63% and 27% respectively compared to a layer-wise approach, for the same accuracy.

Neural Architecture Search (NAS) is increasingly popular to automatically explore the accuracy versus computational complexity trade-off of Deep Learning (DL) architectures. When targeting tiny edge devices, the main challenge for DL deployment is matching the tight memory constraints, hence most NAS algorithms consider model size as the complexity metric. Other methods reduce the energy or latency of DL models by trading off accuracy and number of inference operations. Energy and memory are rarely considered simultaneously, in particular by low-search-cost Differentiable NAS (DNAS) solutions. We overcome this limitation proposing the first DNAS that directly addresses the most realistic scenario from a designer's perspective: the co-optimization of accuracy and energy (or latency) under a memory constraint, determined by the target HW. We do so by combining two complexity-dependent loss functions during training, with independent strength. Testing on three edge-relevant tasks from the MLPerf Tiny benchmark suite, we obtain rich Pareto sets of architectures in the energy vs. accuracy space, with memory footprints constraints spanning from 75% to 6.25% of the baseline networks. When deployed on a commercial edge device, the STM NUCLEO-H743ZI2, our networks span a range of 2.18x in energy consumption and 4.04% in accuracy for the same memory constraint, and reduce energy by up to 2.2x with negligible accuracy drop with respect to the baseline., Comment: Accepted for publication at the ISLPED 2022 ACM/IEEE International Symposium on Low Power Electronics and Design

Random Forests (RFs) are widely used Machine Learning models in low-power embedded devices, due to their hardware friendly operation and high accuracy on practically relevant tasks. The accuracy of a RF often increases with the number of internal weak learners (decision trees), but at the cost of a proportional increase in inference latency and energy consumption. Such costs can be mitigated considering that, in most applications, inputs are not all equally difficult to classify. Therefore, a large RF is often necessary only for (few) hard inputs, and wasteful for easier ones. In this work, we propose an early-stopping mechanism for RFs, which terminates the inference as soon as a high-enough classification confidence is reached, reducing the number of weak learners executed for easy inputs. The early-stopping confidence threshold can be controlled at runtime, in order to favor either energy saving or accuracy. We apply our method to three different embedded classification tasks, on a single-core RISC-V microcontroller, achieving an energy reduction from 38% to more than 90% with a drop of less than 0.5% in accuracy. We also show that our approach outperforms previous adaptive ML methods for RFs., Comment: Published in: 2021 IFIP/IEEE 29th International Conference on Very Large Scale Integration (VLSI-SoC), 2021

Human Activity Recognition (HAR) is a relevant inference task in many mobile applications. State-of-the-art HAR at the edge is typically achieved with lightweight machine learning models such as decision trees and Random Forests (RFs), whereas deep learning is less common due to its high computational complexity. In this work, we propose a novel implementation of HAR based on deep neural networks, and precisely on Binary Neural Networks (BNNs), targeting low-power general purpose processors with a RISC-V instruction set. BNNs yield very small memory footprints and low inference complexity, thanks to the replacement of arithmetic operations with bit-wise ones. However, existing BNN implementations on general purpose processors impose constraints tailored to complex computer vision tasks, which result in over-parametrized models for simpler problems like HAR. Therefore, we also introduce a new BNN inference library, which targets ultra-compact models explicitly. With experiments on a single-core RISC-V processor, we show that BNNs trained on two HAR datasets obtain higher classification accuracy compared to a state-of-the-art baseline based on RFs. Furthermore, our BNN reaches the same accuracy of a RF with either less memory (up to 91%) or more energy-efficiency (up to 70%), depending on the complexity of the features extracted by the RF., Comment: Published in: 2021 18th ACM International Conference on Computing Frontiers (CF)

Energy-efficient machine learning models that can run directly on edge devices are of great interest in IoT applications, as they can reduce network pressure and response latency, and improve privacy. An effective way to obtain energy-efficiency with small accuracy drops is to sequentially execute a set of increasingly complex models, early-stopping the procedure for "easy" inputs that can be confidently classified by the smallest models. As a stopping criterion, current methods employ a single threshold on the output probabilities produced by each model. In this work, we show that such a criterion is sub-optimal for datasets that include classes of different complexity, and we demonstrate a more general approach based on per-classes thresholds. With experiments on a low-power end-node, we show that our method can significantly reduce the energy consumption compared to the single-threshold approach., Comment: Published in: 2020 27th IEEE International Conference on Electronics, Circuits and Systems (ICECS)