EPISODE · May 25, 2025 · 11 MIN
Compressed Federated Learning of Tiny Language Models
from Neural intel Pod · host Neuralintel.org
This document details research into improving Federated Learning (FL) efficiency in autonomous mobile networks by incorporating tiny language models (TLMs) for predicting network performance features. It focuses on the challenge of communication overhead in FL due to frequent neural network data exchanges. The paper proposes and evaluates the use of NNCodec, an implementation of the ISO/IEC Neural Network Coding (NNC) standard, to compress the data shared between participating network cells (clients) and a central server. Experimental results using the Berlin V2X dataset demonstrate that NNCodec can significantly reduce communication data while maintaining the prediction performance of the TLMs.
What this episode covers
This document details research into improving Federated Learning (FL) efficiency in autonomous mobile networks by incorporating tiny language models (TLMs) for predicting network performance features. It focuses on the challenge of communication overhead in FL due to frequent neural network data exchanges. The paper proposes and evaluates the use of NNCodec, an implementation of the ISO/IEC Neural Network Coding (NNC) standard, to compress the data shared between participating network cells (clients) and a central server. Experimental results using the Berlin V2X dataset demonstrate that NNCodec can significantly reduce communication data while maintaining the prediction performance of the TLMs.
NOW PLAYING
Compressed Federated Learning of Tiny Language Models
No transcript for this episode yet
Similar Episodes
Mar 14, 2026 ·23m
Mar 11, 2026 ·16m
Feb 28, 2026 ·14m