Results 71 to 80 of about 9,649,421 (248)

starU‐Net: An enhanced U‐Net architecture with star operation and multi‐view fusion for improved vessel segmentation

VIEW, EarlyView.
Low‐contrast capillaries are often missed or fragmented by segmentation models, so we developed a shallow four‐level starU‐Net that combines star operation‐based feature extraction with dynamic snake convolution‐based multi‐view fusion for improved thin vessel continuity.
Mengwei Bai, Li Tong, Jing Lin, Peng Huang   +3 more
wiley   +1 more source

Batched Low-Rank Adaptation of Foundation Models [PDF]

Low-Rank Adaptation (LoRA) has recently gained attention for fine-tuning foundation models by incorporating trainable low-rank matrices, thereby reducing the number of trainable parameters.
Wen, Yeming, Chaudhuri, Swarat
core   +1 more source

SA-LoRA: Shared-A decoupled low-rank adaptation for class-incremental learning

Journal of King Saud University: Computer and Information Sciences
Parameter-efficient fine-tuning methods have shown promise for continual learning with pre-trained models, yet existing approaches either sacrifice performance or incur linear parameter growth with task count.
Xiaohuan Bing, Yuanhao Xi, Ramin Yahyapour   +2 more
doaj   +1 more source

Exploring a New Architecture for Efficient Parameter Fine-Tuning in SLoRA Multitasking Scenarios

Applied Sciences
Propose an enhanced LoRA (Low-Rank Adaptation) MoE (mixed expert) architecture, SLoRA (Enhanced LoRA MoE Architecture), aimed at addressing the key problem of efficient parameter fine-tuning in multitasking scenarios.
Ce Shi, Jin-Woo Jung
doaj   +1 more source

Enabling pandemic‐resilient healthcare: Narrowband Internet of Things and edge intelligence for real‐time monitoring

CAAI Transactions on Intelligence Technology, EarlyView.
Abstract The Internet of Things (IoT) in deploying robotic sprayers for pandemic‐associated disinfection and monitoring has garnered significant attention in recent research. The authors introduce a novel architectural framework designed to interconnect smart monitoring robotic devices within healthcare facilities using narrowband Internet of Things ...
Md Motaharul Islam, Mohammad Kamrul Hasan, Shayla Islam, Mohammed Balfaqih, Ahmed Ibrahim Alzahrani, Nasser Alalwan, Nurhizam Safie, Zaheed Ahmed Bhuiyan, Rahul Thakkar, Taher M. Ghazal   +9 more
wiley   +1 more source

Janus-LoRA: A Balanced Low-Rank Adaptation for Continual Learning

Low-Rank Adaptation (LoRA) has emerged as a promising paradigm for Continual Learning. It independently updates its low-rank factors ($A$ and $B$), creating a composite update to the full weight matrix through their interaction. To prevent catastrophic forgetting, this update should remain orthogonal to the task-specific subspace that contains ...
Chen, Cheng, Zeng, Pengpeng, Guo, Yuyu, Gao, Lianli, Shen, Hengtao, Song, Jingkuan   +5 more
openaire   +2 more sources

ABM-LoRA: Activation Boundary Matching for Fast Convergence in Low-Rank Adaptation

CoRR
16 pages, 5 figures, under ...
Dongha Lee, Jinhee Park 0001, Minjun Kim, Junseok Kwon   +3 more
openaire   +3 more sources

PowerGPT‐R1: Vision‐Language Reinforcement Fine‐Tuning With Verifiable Reward for Power Inspection

High Voltage, EarlyView.
ABSTRACT Vision‐driven intelligent power inspection systems have long faced the challenge of scarce high‐quality datasets in specialised domains, leading to limited performance of traditional deep learning methods (e.g., Faster‐RCNN) in few‐shot learning scenarios. Recent breakthroughs in language models, particularly the open‐source DeepSeek‐R1 model,
Yangyang Zhong, Pengxin Luo, Yunfeng Yan, Jun Feng, Shixiang Tang, Weizhen He, Yiheng Deng, Yi Chen, Yuhao Zhou, Qi Li, Xianbo Wang, Fushuan Wen, Donglian Qi   +12 more
wiley   +1 more source

Transforming Applied Behavior Analysis Therapy: An Internet of Things-Guided, Retrieval-Augmented Large Language Model Framework

IEEE Access
We propose ABA-RAG, a retrieval-augmented generation (RAG) framework specifically tailored for applied behavior analysis (ABA) interventions, which integrates real-time emotional and behavioral data from Internet-of-Things (IoT) wearable devices.
Haomin Qi, Sin Chung-Ho, Rosanna Yuen-Yan Chan, Chun Man Victor Wong   +3 more
doaj   +1 more source

ID-LoRA: Efficient Low-Rank Adaptation Inspired by Matrix Interpolative Decomposition

CoRR
LoRA has become a universal Parameter-Efficient Fine-Tuning (PEFT) technique that equips Large Language Models (LLMs) to adapt quickly to new tasks. However, when these models are scaled up, even the latest LoRA variants still introduce considerable overhead in trainable parameters.
Xindian Ma, Rundong Kong, Peng Zhang 0002, Ruoxiang Huang, Yongyu Jiang   +4 more
openaire   +2 more sources