دکتر سعید حسینی

Sohar University

sahosseini@su.edu.om

صفحه شخصی

دکتر عبدالرضا ابهری

Toronto Metropolitan University

aabhari@cs.torontomu.ca

صفحه شخصی

دکتر نعمت الله شیری

Concordia Unversity

shiri@cse.concordia.ca

صفحه شخصی

From Value-wise ConvNets to Signal-wise Transformers: Toward LLM-inspired Architectures for Vectorized Data

The study explores adapting Large Language Models (LLMs), particularly transformer-based architectures, to process continuous signal data. It proposes representing signals as vectorized tokens input into transformers, redesigning core components like embeddings, self-attention, and feedforward layers for efficiency and effectiveness, and using the model’s output head to generate meaningful vector representations for signal processing applications.

Empowering Digital Twins with AI Generative Models

Digital twins are virtual representations of physical systems that reflect real-time behavior, performance, and status. Powered by sensor networks, these twins receive continuous data streams that mirror real-world conditions. Integrating AI generative models enables the creation of synthetic data and simulation of complex scenarios, enhancing prediction, optimization, and decision-making capabilities within the digital twin framework.This work explores the integration of AI generative models with digital twins—virtual counterparts of physical systems that rely on sensor networks for real-time data input.