AI Efficacy in Sparse Data Environments: Exploring Approximate Knowledge Interpolation for Practical Applications

Research output: Contribution to journalArticlepeer-review

Abstract

AI stands at the forefront of transforming global industries, achieving remarkable progress in recent years, largely driven by advanced deep learning techniques adept at processing extensive datasets. However, a crucial question arises when confronted with limited and ambiguously characterised data for a novel problem: Can AI maintain its effectiveness under such constraints? This presentation delves into addressing this query, emphasising the role of Fuzzy Rule Interpolation (FRI) in enabling approximate reasoning amidst sparse or incomplete knowledge. This becomes particularly crucial when traditional rule-based inference mechanisms struggle due to misalignment with observations. Extensive research into FRI techniques within computational intelligence has yielded various methodologies. The focus of this presentation centres on a notable subset, Transformation-based FRI (T-FRI). T-FRI operates by mathematically adjusting rules that share similarities with unmatched observations, utilising linear transformations of the nearest rules chosen automatically relative to an unmatched observation.
Original languageEnglish
JournalInternational Journal of Simulation: Systems, Science and Technology
Volume25
Issue number1
DOIs
Publication statusE-pub ahead of print - 05 May 2024

Fingerprint

Dive into the research topics of 'AI Efficacy in Sparse Data Environments: Exploring Approximate Knowledge Interpolation for Practical Applications'. Together they form a unique fingerprint.

Cite this