Adaptive Semantic Token Selection for AI-native Goal-oriented Communications

arXiv:2405.02330v1 Announce Type: cross
Abstract: In this paper, we propose a novel design for AI-native goal-oriented communications, exploiting transformer neural networks under dynamic inference constraints on bandwidth and computation. Transformers have become the standard architecture for pretraining large-scale vision and text models, and preliminary results have shown promising performance also in deep joint source-channel coding (JSCC). Here, we consider a dynamic model where communication happens over a channel with variable latency and bandwidth constraints. Leveraging recent works on conditional computation, …

abstract ai-native architecture arxiv bandwidth become communications computation constraints cs.ai cs.it cs.lg design dynamic inference math.it networks neural networks novel paper performance pretraining results scale semantic standard text token transformer transformers type vision