Knowledge Distillation of Uncertainty using Deep Latent Factor Model

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Park, Sehyun, Lee, Jongjin, Shin, Yunseop, Ohn, Ilsang, Kim, Yongdai
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866909866972413952
author Park, Sehyun
Lee, Jongjin
Shin, Yunseop
Ohn, Ilsang
Kim, Yongdai
author_facet Park, Sehyun
Lee, Jongjin
Shin, Yunseop
Ohn, Ilsang
Kim, Yongdai
contents Deep ensembles deliver state-of-the-art, reliable uncertainty quantification, but their heavy computational and memory requirements hinder their practical deployments to real applications such as on-device AI. Knowledge distillation compresses an ensemble into small student models, but existing techniques struggle to preserve uncertainty partly because reducing the size of DNNs typically results in variation reduction. To resolve this limitation, we introduce a new method of distribution distillation (i.e. compressing a teacher ensemble into a student distribution instead of a student ensemble) called Gaussian distillation, which estimates the distribution of a teacher ensemble through a special Gaussian process called the deep latent factor model (DLF) by treating each member of the teacher ensemble as a realization of a certain stochastic process. The mean and covariance functions in the DLF model are estimated stably by using the expectation-maximization (EM) algorithm. By using multiple benchmark datasets, we demonstrate that the proposed Gaussian distillation outperforms existing baselines. In addition, we illustrate that Gaussian distillation works well for fine-tuning of language models and distribution shift problems.
format Preprint
id arxiv_https___arxiv_org_abs_2510_19290
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Knowledge Distillation of Uncertainty using Deep Latent Factor Model
Park, Sehyun
Lee, Jongjin
Shin, Yunseop
Ohn, Ilsang
Kim, Yongdai
Machine Learning
Methodology
Deep ensembles deliver state-of-the-art, reliable uncertainty quantification, but their heavy computational and memory requirements hinder their practical deployments to real applications such as on-device AI. Knowledge distillation compresses an ensemble into small student models, but existing techniques struggle to preserve uncertainty partly because reducing the size of DNNs typically results in variation reduction. To resolve this limitation, we introduce a new method of distribution distillation (i.e. compressing a teacher ensemble into a student distribution instead of a student ensemble) called Gaussian distillation, which estimates the distribution of a teacher ensemble through a special Gaussian process called the deep latent factor model (DLF) by treating each member of the teacher ensemble as a realization of a certain stochastic process. The mean and covariance functions in the DLF model are estimated stably by using the expectation-maximization (EM) algorithm. By using multiple benchmark datasets, we demonstrate that the proposed Gaussian distillation outperforms existing baselines. In addition, we illustrate that Gaussian distillation works well for fine-tuning of language models and distribution shift problems.
title Knowledge Distillation of Uncertainty using Deep Latent Factor Model
topic Machine Learning
Methodology
url https://arxiv.org/abs/2510.19290