Mean-Field-Type Transformers

Chapman & Hall/CRC eBooks 2026 Book chapter

View at publisher
Authors: Hamidou Tembiné
DOI: 10.1201/9781003664581-4
Cited by: 0

Abstract

This chapter introduces mean-field-type transformers. Attention depends on both individual states and population distributions. New architectures are constructed explicitly. Difference attention and HoloNorm are analyzed. Outcomes are characterized as mean-field-type games. Training is reformulated accordingly. Diffusion and time-reversed formulations are developed. Federated learning and unlearning are integrated. Constant parameters are proven suboptimal. Post-training adaptation is formalized without retraining. The chapter unifies architecture, learning, and mean-field effects.

Fields of Study & Subjects

Neural Networks and Applications Theoretical and Computational Physics Advanced DC-DC Converters

Publication details

Journal / source
Chapman & Hall/CRC eBooks
Pages
113–174
Publication date
6 February 2026
Language
EN
Document type
Book chapter