187037 research outputs found
Sort by
Low-Power Power Management Unit With Adaptive Dynamic Load Power Tracking for Millimeter-Scale Computing Systems
No full textThis article proposes a power management unit (PMU) for a millimeter-scale, multilayer-stacked computing system, designed to efficiently manage significant load current variations between low-power sleep mode and high-performance active mode. The system ensures a reliable power supply across different chip layers, rapidly increasing output driving strength in response to active-mode requests from any layers. In addition, the circuit adaptively manages the switching frequency of its power converters for dynamic frequency scaling, compensating for temperature changes by using an oscillator that replicates the processor's clock. The proposed design is fabricated using a 180-nm process and integrated with other chip layers in a millimeter-scale system. The system's PMU eliminates voltage drop during transitions to active mode by utilizing a constant energy-per-cycle oscillator with a wide frequency range of 1.89 Hz-104 MHz, along with a replica oscillator of the processor. It achieves over 50% power conversion efficiency across a load power range of 41 nW-409 mu W.
AIPO: Automatic Instruction Prompt Optimization by model itself with "Gradient Ascent
No full textLarge language models (LLMs) can perform a variety of tasks such as summarization, translation, and question answering by generating answers with user input prompt. The text that is used as input to the model, including instruction, is called input prompt. There are two types of input prompt: zero-shot prompting provides a question with no examples, on the other hand, few-shot prompting provides a question with multiple examples. The way the input prompt is set can have a big impact on the accuracy of the model generation. The relevant research is called prompt engineering. Prompt engineering, especially prompt optimization is used to find the optimal prompts optimized for each model and task. Manually written prompts could be optimal prompts, but it is time-consuming and expensive. Therefore, research is being conducted on automatically generating prompts that are as effective as human-crafted ones for each task. We propose Automatic Instruction Prompt Optimization (AIPO), which allows the model to generate an initial prompt directly through instruction induction when given a task in a zero-shot setting and then improve the initial prompt to optimal prompt for model based on the "gradient ascent" algorithm. With the final prompt generated by AIPO, we achieve more accurate generation than manual prompt on benchmark datasets regardless of the output format.
On-demand, reversible, and reusable finger-actuated air pumping module for fully integrated modular microfluidics
No full textModular microfluidics can reduce the time and cost of microfluidic chip development; however, reliance on commercial pumping systems limits widespread use for prototyping. In this study, we present a modular pump, a finger-actuated air pumping module (FAPMO), which enables on-demand, repeated, reversible, and versatile operation of modular microfluidics. The FAPMO features integrated microvalves with a pushbutton that enables repeated operations in both expulsion and suction modes, and coupling the FAPMO with a reservoir module facilitates versatile use. By integrating a polydimethylsiloxane (PDMS)-based pumping component into a modular cage with interfacing film channels, the FAPMO applies pressure to other modules. After characterizing the pressure driven by the FAPMO and the flow velocity of a reagent from the reservoir module, we demonstrate its application in the purification of bacterial DNA. By assembling and operating modular blocks for reagent injection, mixing, washing, and elution, we successfully purified DNA from a lysate at concentrations as low as 104 CFU/mL of Escherichia coli O157:H7. The FAPMO is expected to be widely applicable in modular microfluidics, enabling straightforward proof of concept in biological and chemical applications.
SECTRACER: A framework for uncovering the root causes of network intrusions via security provenance
No full textModern enterprise networks comprise diverse and heterogeneous systems that support a wide range of services, making it challenging for administrators to track and analyze sophisticated attacks such as advanced persistent threats (APTs), which often exploit multiple vectors. To address this challenge, we introduce the concept of network-level security provenance, which enables the systematic establishment of causal relationships across hosts at the network level, facilitating the accurate identification of the root causes of security incidents. Building on this concept, we present SECTRACER as a framework for a network-wide provenance analysis. SECTRACER offers three main contributions: (i) comprehensive and efficient forensic data collection in enterprise networks via software-defined networking (SDN), (ii) reconstruction of attack histories through provenance graphs to provide a clear and interpretable view of intrusions, and (iii) proactive attack prediction using probabilistic models. We evaluated the effectiveness and efficiency of SECTRACER through a real-world APT simulation, demonstrating its capability to enhance threat mitigation while introducing less than 1 % network throughput overhead and negligible latency impact.
1,4-Naphthoquinone improves depressive-like behaviors by modulating neuronal factors and neuroinflammatory mediators
No full textDepression is a chronic mental disorder characterized by alternations in emotions, thoughts, physical condition, and behavior. Using the natural product database Compound Combination-Oriented Natural Product Database with Unified Terminology (COCONUT) and the bioinformatics tool CODA (Context-Oriented Directed Associations), we screened and identified 1,4-naphthoquinone (1,4-NQ) as a promising candidate for depression treatment. Oral administration of 1,4-NQ attenuated the depressive-like behaviors in the open field test (OFT), elevated plus maze test (EPM) and forced swim test (FST) in a chronic restraint stress (CRS)-induced depressive-like mouse model. Real-time PCR analysis demonstrated that 1,4-NQ increased the mRNA levels of 5-HT1A and BDNF in the hippocampus of mouse brains. The expression level of glucocorticoid receptor (GR) in the hippo-campus was increased by 1,4-NQ treatment in both CRS-and corticosterone-induced depression mouse models. We confirmed that 1,4-NQ has anti-neuroinflammatory efficacy by suppressing the levels of IL-6, TNF-alpha and IL-1 beta in LPS-stimulated BV2 microglial cell line. The western blot and real-time PCR analysis demonstrated that 1,4-NQ increased the level of GR in both the U-138 MG glial cell line and the SH-SY5Y neuronal cell line. In conclusion, 1,4-NQ is supposed to have anti-depressive efficacy by alleviating depressive-like behaviors through modulation of neuroinflammatory mediators and GR expression in the nervous system.
Effect of soil type on effective soil thermal conductivity for the full range of water saturation
No full textAccurate estimation of effective soil thermal conductivity is crucial for designing and managing geosystems such as underground power cables, thermally active geostructures, and nuclear waste repositories. Previous studies have examined the soil thermal response by varying soil properties and water saturation for specific soil types. However, the applicability of these methods in capturing the evolution of effective soil thermal conductivity across different soil types and fundamental soil properties is limited. This study investigates the effect of soil type and fundamental soil properties on effective soil thermal conductivity across the full range of water saturation. A predictive model is developed to describe the evolution of normalized thermal conductivity with saturation, incorporating two physically meaningful parameters that characterize the initial and intermediate thermal response with increasing water saturation. This model adequately fits the effective soil thermal conductivity data collected from the literature. Using an extensive dataset of various soil types, soils are classified into three major texture groups: coarse-textured, moderately coarse-to medium-textured, and moderately fine-to fine-textured soils. These groups exhibit distinct trends in thermal conductivity evolution with water saturation, as reflected in their model parameters. Further analysis explores the key soil properties that govern these model parameters, providing a comprehensive understanding of the mechanisms controlling thermal conductivity variation. As an engineering application, a practical reference is proposed to categorize soils based on key parameters such as the saturated and dry thermal conductivity, initial porosity, and average thermal conductivity of soil particles.
Improved query specialization for transformer-based visual relationship detection
No full textVisual Relationship Detection (VRD) has significantly advanced with Transformer-based architectures. However, we identify two fundamental drawbacks in conventional label assignment methods used for training Transformer-based VRD models, where ground-truth (GT) annotations are matched to model predictions. In conventional assignment, queries are trained to detect all relations rather than specializing in specific ones, resulting in 'unspecialized' queries. Also, each ground-truth (GT) annotation is assigned to only one prediction under conventional assignment, suppressing other near-correct predictions by labeling them as 'no relation'. To address these issues, we introduce a novel method called Groupwise Query Specialization and Quality-Aware Multi-Assignment (SpeaQ). Groupwise Query Specialization clusters queries and relations into exclusive groups, promoting specialization by assigning a set of relations only to a corresponding query group. Quality-Aware Multi-Assignment enhances training signals by allowing multiple predictions closely matching the GT to be positively assigned. Additionally, we introduce dynamic query reallocation, which transfers queries from high-to low-performing groups for balanced training. Experimental results demonstrate that SpeaQ+, combining SpeaQ with dynamic query reallocation, consistently improves performance across seven baseline models on five benchmarks without additional inference cost.
Extended modular functions and definite form class groups
No full textFor a positive integer N, we define an extended modular function of level N motivated by physics and investigate its fundamental properties. Let K be an imaginary quadratic field, and let O be an order in K of discriminant D. Let KO, N denote the ray class field of O modulo NO. For N >= 3, we provide an explicit description of the Galois group Gal(KO, N/Q) using special values of extended modular functions of level Nand the definite form class group of discriminant D and level N. (c) 2025 Elsevier Inc. All rights are reserved, including those for text and data mining, AI training, and similar technologies.
Comprehensive assessment of hydrogen jet dynamics under split-injection conditions
No full textIn stratified charge combustion (SCC) approach, direct injection requires optimal air-fuel mixing to maximize engine performance. This study examined the potential of split injection strategy to enhance hydrogen mixing process through both experimental and computational approaches using direct injection hollow-cone injector. Quantitative hydrogen concentration was measured using Laser-induced breakdown spectroscopy (LIBS), while computational fluid dynamics (CFD) simulation was performed to investigate aerodynamics during injection process. The results showed that the first injection induces a pressure difference across hydrogen jet, boundary generating an upward-directed vortex. A shorter first injection duration intensifies the vortical structure in the jet, promoting stronger interaction between the first and second injections. This interaction enhances jet dispersion and improves mixture homogeneity.
Multiscale hybrid wick: Two types of dry-out mechanisms and optimization strategy
No full textCapillary-driven transport in porous wicks are core elements in phase-change-based thermal management solutions such as vapor chambers, cold plates, heat pipes, and heat exchangers. The biggest challenge in developing high performance wicks is the conflict between the heat transfer coefficient and critical heat flux. To address these challenges, hybrid wicks composed of liquid supply wicks connected to thin evaporator wicks have been suggested. Nevertheless, the understanding of the dry-out mechanism of multiscale hybrid wicks was insufficient, which limits the development of thermal-hydraulic performance prediction models. In this work, the dry-out types of multiscale hybrid wick (MSHW) were classified into two categories, and a numerical model reflecting the dry-out mechanism was developed. The two types of dry-out mechanisms and the developed model were validated by capillary-rise and evaporative heat transfer experiments. The design criterion determining the type of dry-out was also suggested. By considering the dry-out mechanisms, the developed model provided 10 times smaller error in CHF prediction compared to that of conventional model. Then the artificial neural networks (ANN) model was trained using the data from the numerical model to efficiently identify the optimal design of the multiscale wicks. The optimized multiscale wick achieved a CHF approximately 24 times higher than that of single scale wicks at a comparable thermal resistance.