Kalman Bayesian Transformer

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Jing, Haoming, Wright, Oren, Moura, José M. F., Nakahira, Yorie
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912585448685568
author Jing, Haoming
Wright, Oren
Moura, José M. F.
Nakahira, Yorie
author_facet Jing, Haoming
Wright, Oren
Moura, José M. F.
Nakahira, Yorie
contents Sequential fine-tuning of transformers is useful when new data arrive sequentially, especially with shifting distributions. Unlike batch learning, sequential learning demands that training be stabilized despite a small amount of data by balancing new information and previously learned knowledge in the pre-trained models. This challenge is further complicated when training is to be completed in latency-critical environments and learning must additionally quantify and be mediated by uncertainty. Motivated by these challenges, we propose a novel method that frames sequential fine-tuning as a posterior inference problem within a Bayesian framework. Our approach integrates closed-form moment propagation of random variables, Kalman Bayesian Neural Networks, and Taylor approximations of the moments of softmax functions. By explicitly accounting for pre-trained models as priors and adaptively balancing them against new information based on quantified uncertainty, our method achieves robust and data-efficient sequential learning. The effectiveness of our method is demonstrated through numerical simulations involving sequential adaptation of a decision transformer to tasks characterized by distribution shifts and limited memory resources.
format Preprint
id arxiv_https___arxiv_org_abs_2509_10695
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Kalman Bayesian Transformer
Jing, Haoming
Wright, Oren
Moura, José M. F.
Nakahira, Yorie
Machine Learning
Artificial Intelligence
Sequential fine-tuning of transformers is useful when new data arrive sequentially, especially with shifting distributions. Unlike batch learning, sequential learning demands that training be stabilized despite a small amount of data by balancing new information and previously learned knowledge in the pre-trained models. This challenge is further complicated when training is to be completed in latency-critical environments and learning must additionally quantify and be mediated by uncertainty. Motivated by these challenges, we propose a novel method that frames sequential fine-tuning as a posterior inference problem within a Bayesian framework. Our approach integrates closed-form moment propagation of random variables, Kalman Bayesian Neural Networks, and Taylor approximations of the moments of softmax functions. By explicitly accounting for pre-trained models as priors and adaptively balancing them against new information based on quantified uncertainty, our method achieves robust and data-efficient sequential learning. The effectiveness of our method is demonstrated through numerical simulations involving sequential adaptation of a decision transformer to tasks characterized by distribution shifts and limited memory resources.
title Kalman Bayesian Transformer
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2509.10695