Timezone: »

Variational Model Perturbation for Source-Free Domain Adaptation
Mengmeng Jing · Xiantong Zhen · Jingjing Li · Cees Snoek

Tue Nov 29 09:00 AM -- 11:00 AM (PST) @ Hall J #425

We aim for source-free domain adaptation, where the task is to deploy a model pre-trained on source domains to target domains. The challenges stem from the distribution shift from the source to the target domain, coupled with the unavailability of any source data and labeled target data for optimization. Rather than fine-tuning the model by updating the parameters, we propose to perturb the source model to achieve adaptation to target domains. We introduce perturbations into the model parameters by variational Bayesian inference in a probabilistic framework. By doing so, we can effectively adapt the model to the target domain while largely preserving the discriminative ability. Importantly, we demonstrate the theoretical connection to learning Bayesian neural networks, which proves the generalizability of the perturbed model to target domains. To enable more efficient optimization, we further employ a parameter sharing strategy, which substantially reduces the learnable parameters compared to a fully Bayesian neural network. Our model perturbation provides a new probabilistic way for domain adaptation which enables efficient adaptation to target domains while maximally preserving knowledge in source models. Experiments on several source-free benchmarks under three different evaluation settings verify the effectiveness of the proposed variational model perturbation for source-free domain adaptation.

Author Information

Mengmeng Jing (University of Electronic Science and Technology of China)

Mengmeng Jing received the B.Eng. and M.Sc. degrees in 2015 and 2018, respectively, from the University of Electronic Science and Technology of China, Chengdu, China, where he is currently working on the Ph.D. degree with the School of Computer Science and Engineering. His research interests include machine learning, especially transfer learning, and subspace learning.

Xiantong Zhen (United Imaging Healthcare)
Jingjing Li (University of Electronic Science and Technology of China)
Cees Snoek (University of Amsterdam)

More from the Same Authors