2022 IEEE International Symposium on Circuits and Systems (ISCAS), Journal Year: 2024, Volume and Issue: unknown, P. 1 - 5
Published: May 19, 2024
Language: Английский
DNA repair, Journal Year: 2023, Volume and Issue: 128, P. 103529 - 103529
Published: June 26, 2023
DNA adducts and strand breaks are induced by various exogenous endogenous agents. Accumulation of damage is implicated in many disease processes, including cancer, aging, neurodegeneration. The continuous acquisition from stressors coupled with defects repair pathways contribute to the accumulation within genome genomic instability. While mutational burden offers some insight into level a cell may have experienced subsequently repaired, it does not quantify breaks. Mutational also infers identity damage. With advances adduct detection quantification methods, there an opportunity identify driving mutagenesis correlate known exposome. However, most methods require isolation or separation its context nuclei. Mass spectrometry, comet assays, other techniques precisely lesion types but lose nuclear even tissue growth spatial analysis technologies novel leverage context. we lack wealth capable detecting situ. Here, review limited existing situ examine their potential offer tumors tissues. We perspective on need for highlight Repair Assisted Damage Detection (RADD) as technique integrate challenges be addressed.
Language: Английский
Citations
7
Electronics, Journal Year: 2024, Volume and Issue: 13(7), P. 1281 - 1281
Published: March 29, 2024
Extensive research has been carried out on reinforcement learning methods. The core idea of is to learn methods by means trial and error, it successfully applied robotics, autonomous driving, gaming, healthcare, resource management, other fields. However, when building solutions at the edge, not only are there challenges data-hungry insufficient computational resources but also difficulty a single method meet requirements model in terms efficiency, generalization, robustness, so on. These rely expert knowledge for design edge-side integrated methods, they lack high-level system architecture support their wider generalization application. Therefore, this paper, instead surveying systems, we survey most commonly used options each part from point view We present characteristics traditional several aspects corresponding integration framework based them. In process, show complete primer architectures while demonstrating flexibility various parts be adapted different edge tasks. Overall, become an important tool intelligent decision making, still faces many practical application computing. aim paper provide researchers practitioners with new, perspective better understand apply decision-making
Language: Английский
Citations
2
Computers & Electrical Engineering, Journal Year: 2024, Volume and Issue: 116, P. 109180 - 109180
Published: April 6, 2024
Language: Английский
Citations
2
Sensors, Journal Year: 2024, Volume and Issue: 24(19), P. 6435 - 6435
Published: Oct. 4, 2024
To create an effective Ambient Assisted Living (AAL) system that supports the daily activities of patients or elderly, it is crucial to accurately detect and differentiate user actions determine necessary assistance. Traditional intrusive methods, such as wearable object-attached devices, can interfere with natural behavior may lead resistance. Furthermore, non-intrusive systems rely on video sound data processed by servers cloud generate excessive traffic raise concerns about security personal information. In this study, we developed edge-based real-time for detecting Activities Daily (ADL) using ambient noise. Additionally, introduced online post-processing method enhance classification performance extract activity events from noisy in resource-constrained environments. The system, tested collected a living space, achieved high accuracy classifying ADL-related behaviors continuous successfully generated logs time-series data, enabling further analyses ADL assessments. Future work will focus enhancing detection expanding range detectable integrating study additional sources beyond sound.
Language: Английский
Citations
2
Discover Sustainability, Journal Year: 2024, Volume and Issue: 5(1)
Published: Nov. 15, 2024
The rapid advancement of artificial intelligence (AI) has brought significant benefits across various domains, yet it also led to increased energy consumption and environmental impact. This paper positions Green AI as a crucial direction for future research development. It proposes comprehensive framework understanding, implementing, advancing sustainable practices. We provide an overview AI, highlighting its significance current state regarding AI's explores techniques, such model optimization methods, the development efficient algorithms. Additionally, we review energy-efficient hardware alternatives like tensor processing units (TPUs) field-programmable gate arrays (FPGAs), discuss strategies designing operating data centers. Case studies in natural language (NLP) Computer Vision illustrate successful implementations Through these efforts, aim balance performance resource efficiency technologies, aligning them with global sustainability goals.
Language: Английский
Citations
2
Scientific Reports, Journal Year: 2024, Volume and Issue: 14(1)
Published: July 16, 2024
Abstract Resource optimization, timely data capture, and efficient unmanned aerial vehicle (UAV) operations are of utmost importance for mission success. Latency, bandwidth constraints, scalability problems the that conventional centralized processing architectures encounter. In addition, optimizing robust communication between ground stations UAVs while protecting privacy security is a daunting task in itself. Employing edge computing infrastructure, artificial intelligence-driven decision-making, dynamic offloading mechanisms, this research proposes edge-aware optimization framework (DTOE-AOF) UAV optimization. Edge intelligence (AI) algorithms integrate to decrease latency, increase efficiency, conserve onboard resources. This system dynamically assigns duties nodes according proximity, available resources, urgency tasks. Reduced increased resource conservation result from implementation (DTOE-AIF)'s integration AI with computing. DTOE-AOF useful many fields, such as precision agriculture, emergency management, infrastructure inspection, monitoring. powered by outfitted can swiftly survey damage, find survivors, launch rescue missions. By comparing methods, thorough simulation confirms it improves response time, utilization.
Language: Английский
Citations
1
SoftwareX, Journal Year: 2024, Volume and Issue: 27, P. 101858 - 101858
Published: Aug. 23, 2024
The application of pruning techniques to convolutional neural networks has made it possible reduce the size model and time required for inference. However, determining best policy, i.e. pair method distribution that allows obtaining highest accuracy or F1 score pruned model, is not a task can be easily performed with available tools. For this, we propose library called FlexiPrune, written in Python language using Pytorch framework, which user select an unpruned choose policy from set options. FlexiPrune makes very easy compare impact different methods distributions, so decision making based on performance specific GPR (Global Pruning Rate) value classification problem, rather than simply following generic recommendations.
Language: Английский
Citations
1
Advances in business strategy and competitive advantage book series, Journal Year: 2024, Volume and Issue: unknown, P. 123 - 150
Published: Aug. 26, 2024
Integrating artificial intelligence (AI) into management science practices has become one of the most important ways to boost productivity in today's business world. This chapter focuses on how AI technologies have changed different parts an organization work. Businesses can improve overall efficiency, make decisions more quickly, and run processes smoothly using AI. The examines core methodologies, including machine learning, natural language processing, predictive analytics, their applications automating routine tasks, personalizing customer interactions, enabling data-driven strategies. They are used automate contacts with customers personal, strategies possible. AI-powered tools changing people work, encouraging new ideas, giving companies edge over competitors through in-depth research real-life case studies. Problems moral issues connected AI, a complete picture will help organizations productive future.
Language: Английский
Citations
1
Asian Journal of Electrical and Electronic Engineering., Journal Year: 2024, Volume and Issue: 4(2), P. 52 - 60
Published: Oct. 2, 2024
As the IoT ecosystem continues to grow, edge computing is becoming essential for handling and analyzing vast amount of data generated by connected devices. Unlike traditional centralized models, where information sent remote centers processing, processes closer it generated. This decentralized approach helps reduce latency, optimizes bandwidth usage, improves both privacy security. However, rise in devices spread also increase potential cyberattacks, demanding more robust security measures. With AI machine learning being utilized analyze data, facilitates this analysis directly at source, pointing a future ML applications are prevalent on
Language: Английский
Citations
1
Published: Nov. 22, 2024
Chapter 4 delves into various model optimization techniques crucial for deploying AI models on edge devices such as smartphones, smartwatches, and IoT devices. These optimizations are categorized three phases: predeployment, deployment-time, postdeployment. Predeployment include architecture selection, quantization, structured pruning, knowledge distillation, sparsification, which applied to the before production enhance performance efficiency. Deployment-time techniques, IR conversion, graph optimizations, target-dependent dynamic batching, caching, parallelism, employed optimize during deployment runtime. Postdeployment including monitoring, retraining, hardware upgrades, user feedback loops, ensure continuous improvement adaptability of in real-world scenarios. Through illustrative examples, this chapter provides a comprehensive understanding how these strategies can be effectively implemented meet constraints requirements computing.
Language: Английский
Citations
1