Convolutional neural networks (CNN) achieve state-of-the-art results in the field of visual perception, drastically changing traditional computer-vision framework. However, movement massive amounts data prevents CNN's from being integrated into low-power IoT devices. The recently proposed binaryweight network (BWN) reduces complexity computation and amount memory access. A conventional digital implementation, which is composed separate feature/weight memories a multiply-and-accumulate (MAC)...

The Google Glass is a mobile device designed to be worn as eyeglasses. This form factor enables new use cases, such hands-free video chat and web search. However, its shape also hampers potential: (1) battery size, therefore lifetime, limited by need for the lightweight, (2) high-power processing leads significant heat, which should due compact proximity user's skin. We an Explorer Edition of (XE12) study power thermal characteristics optical head-mounted display devices. share insights...

Abstract An artificial vision system that can simulate the visual functions of human eyes is required for biological robots. Here, In‐Ga‐Zn‐O memtransistors using a naturally oxidized Al 2 O 3 and an ion gel as common gate stacking dielectric proposed. Positive charge trapping in layer be induced by modulating voltage, which causes back sweep subthreshold swing (SS) device to break physical limit (≥60 mV per decade at room temperature), minimum SS low 26.4 decade. In addition, photogenerated...

Emerging non-volatile memory-based computing-in-memory (CIM) is an excellent fit for resource-constrained edge-AI devices [1–6]. MRAM-CIM macros MAC operations, at present, rely on a crossbar structure or peripheral circuit modification [2], [3]. It remains great challenge bottom-up design of macro using the standard one transistor - magnetic tunnel junction (1T-1MTJ) bit-cell: (1) The mainstream spin-transfer-torque (STT) switching mechanism with foundry bit-cell cannot fulfill CIM...

Abstract The human memory system plays an indispensable role in oblivion, learning, and memorization. Implementing a within electronic devices contributes important step toward constructing neuromorphic that emulates advanced mental functions of the brain. Given complex time‐tailoring requirement, integrating into one is great challenge. Here, van der Waals heterostructure with flexible ability demonstrated, which can meet high requirement programming. By stacking volatile nonvolatile...

As automated scoring systems for spoken responses are increasingly used in language assessments, testing organizations need to analyze their performance, as compared human raters, across several dimensions, example, on individual items or based subgroups of test takers. In addition, there is a establish rigorous procedures monitoring the performance both and processes during operational administrations. This paper provides an overview speech system SpeechRater SM how use charts evaluation...

This paper proposes a high power-performance-area efficient background noise aware keyword-spotting (KWS) processor based on an optimized binarized weight network (BWN). To reduce the power consumption while maintaining system recognition accuracy for different noise, KWS with SNR prediction module can be adaptively configured to use dual computing modes (standard mode and approximate mode) both under ultra-low low noise. The mel-scale frequency cepstral coefficients (MFCC) is technologies,...

To accomplish frequent, simple tasks with high efficiency, it is necessary to leverage low-power, microcontroller-like processors that are increasingly available on mobile systems. However, existing solutions require developers directly program the low-power and carefully manage inter-processor communication. We present Reflex, a suite of compiler runtime techniques significantly lower barrier for such processors. The heart Reflex software Distributed Shared Memory (DSM) enables shared...

The exchange bias (EB) effect denotes a magnetic phenomenon originating from the interfacial coupling at ferromagnetic/antiferromagnetic materials, which plays an indispensable role in functionality of various devices, such as random-access memory (MRAM) and sensors. Voltage control offers promising pathway to significantly reduce device power consumption, effectively fostering evolution low-energy spintronic devices. “magneto-ionic” mechanism, characterized by its operational efficiency,...

The early warning of disasters such as ground pressure in deep hard rock mines has long constrained the safe and efficient development mining activities. Based on fractal theory dimension interpretation, this study constructs a microseismic monitoring system for areas, extracting key elements, particularly time energy elements. Using box-counting method theory, investigates dimensions time–energy data disaster source warning. Through parameter analysis, events related to local potential...

Recent advances in text-to-image diffusion models have shown an outstanding ability zero-shot style transfer. However, existing methods often struggle to balance preserving the semantic content of input image and faithfully transferring target line with edit prompt. Especially when applied complex traffic scenes diverse objects, layouts, stylistic variations, current tend exhibit Style Neglection, i.e., failing generate required To address this issue, we propose Nursing, which directs model...

This paper presents system-architecture-circuits co-designs for computing the MFCC feature extraction speech keywords recognition. The trade-off between accuracy and power consumption under various background noises is achieved by using 8-stage radix-2 single-path delay feedback FFT (R2SDF-FFT) precision self-adaptive architecture with approximate computing. R2SDF-FFT structure fine-grained bit-width quantization can reduce 35.7% of memory size. Approximate multiplication addition Dual-Vdd...

An ultra-low power always-on keyword spotting (KWS) accelerator is implemented in 22nm CMOS technology, which based on an optimized convolutional neural network (CNN). To reduce the consumption while maintaining system recognition accuracy, we first perform a bit-width quantization method proposed CNN to data/weight bit width required by hardware computing unit without reducing accuracy. Then, propose approximate architecture for quantized using voltage-domain analog switching multiplication...

This paper proposed an energy-efficient reconfigurable DNN accelerator architecture for voice activity detection (VAD) based on deep neural networks and fabricated in 28-nm technology. To reduce the power consumption achieve high energy efficiency, two optimization techniques are proposed. First, processing elements contained support digital-analog mixed approximate computing, including multi-step quantized multiplication units time-delay addition units. Second, computing can be dynamically...

Low-power keywords recognition has been a focus of acoustic signal processing for several decades. This work investigates the domain-specific speech microprocessor based on optimized one-dimensional convolutional recurrent neural network (1D-CRNN). Compared to previous DNN frameworks, proposed 1D-CRNN can process both feature extraction and classification, achieve high accuracy with reduced computation operations under wide range background noise SNRs. An energy-efficient accelerator is...

This paper proposed an energy-efficient reconfigurable accelerator for keyword spotting (EERA-KWS) based on binary weight network (BWN) and fabricated in 28-nm CMOS technology. system consists of two parts: the feature extraction melscale frequency cepstral coefficients (MFCC) keywords classification a BWN model, which is trained through Google's Speech Commands database deployed our custom. To reduce power consumption while maintaining recognition accuracy, we first optimize MFCC...

Source detection in graphs has demonstrated robust efficacy the domain of rumor source identification. Although recent solutions have enhanced performance by leveraging deep neural networks, they often require complete user data. In this paper, we address a more challenging task, with incomplete data, and propose novel framework, i.e., Detection Graphs Incomplete Nodes via Positional Encoding Attentive Fusion (GIN-SD), to tackle challenge. Specifically, our approach utilizes positional...

Background: Spontaneous coronary artery dissection (SCAD) is a rare and often underdiagnosed cause of acute syndrome (ACS), predominantly affecting younger women without traditional cardiovascular risk factors. The management SCAD remains subject debate, likely secondary to inconclusive evidence. This study aims compare the clinical outcomes patients treated with optimal medical therapy (OMT) versus those who underwent percutaneous intervention (PCI) using national population-based cohort....

Machine Learning methods have been adopted for a wide range of real-world applications, ranging from social networks, online image/video-sharing platforms, and e-commerce to education, healthcare, etc. However, in practice, large amount effort is required tune several components machine learning methods, including data representation, hyperparameter, model architecture, order achieve good performance. To alleviate the tunning efforts, Automated (AutoML), which can automate process applying...

Studies using natural language processing (NLP) techniques are increasingly being published. Evidence-based medicine (EBM) users need to learn the basics of NLP be able appraise these types studies. We propose a set criteria evaluate quality studies that have used NLP, focusing on methods sample selection, coding, gold standard, algorithm training, testing and measures accuracy (such as recall precision). has proven critical for conducting biomedical research potential improve healthcare...

Although stance detection has made great progress in the past few years, it is still facing problem of unseen targets. In this study, we investigate domain difference between targets and thus incorporate attention-based conditional encoding with adversarial generalization to perform target detection. Experimental results show that our approach achieves new state-of-the-art performance on SemEval-2016 dataset, demonstrating importance