Scholarlink

<p>Cognitive Internet of Things (CIoT) integrates intelligence into IoT infrastructures to convert massive, heterogeneous data streams into actionable knowledge. Processing such data in real time presents significant computational challenges that cannot be addressed without high performance computing (HPC). Therefore, we propose a brain-inspired fluid intelligence framework optimized for HPC environments, featuring a novel irreflexive reasoning mechanism. The approach models datasets as mathematical relations, extracts irreflexive pairs, and computes probabilistic strengths to identify the most significant relationships, all without relying on any training datasets. These distilled relational units enable high computational efficiency in knowledge discovery, forecasting, and recommendation tasks. The framework accelerates relational analysis and probabilistic computation, achieving remarkable scalability for large-scale deployments. Experiments on environmental datasets demonstrate superior accuracy (&gt; 99%), runtime performance, and scalability compared to state-of-the-art methods, highlighting the synergy between cognition-inspired intelligence and HPC for next-generation CIoT applications.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Fluid intelligence: a novel strategy for adaptive learning in cognitive IoT

  • 期刊论文

  • 发表日期：

    2025年11月20日

查看全文
  • Vidyapati Jha,
  • Priyanka Tripathi

摘要

Cognitive Internet of Things (CIoT) integrates intelligence into IoT infrastructures to convert massive, heterogeneous data streams into actionable knowledge. Processing such data in real time presents significant computational challenges that cannot be addressed without high performance computing (HPC). Therefore, we propose a brain-inspired fluid intelligence framework optimized for HPC environments, featuring a novel irreflexive reasoning mechanism. The approach models datasets as mathematical relations, extracts irreflexive pairs, and computes probabilistic strengths to identify the most significant relationships, all without relying on any training datasets. These distilled relational units enable high computational efficiency in knowledge discovery, forecasting, and recommendation tasks. The framework accelerates relational analysis and probabilistic computation, achieving remarkable scalability for large-scale deployments. Experiments on environmental datasets demonstrate superior accuracy (> 99%), runtime performance, and scalability compared to state-of-the-art methods, highlighting the synergy between cognition-inspired intelligence and HPC for next-generation CIoT applications.