Unsupervised data imputation via matrix-variate variational autoencoders

Shenfen Kuang

doi:10.1117/12.3037941

7 August 2024 Unsupervised data imputation via matrix-variate variational autoencoders

Shenfen Kuang

Proceedings Volume 13229, Seventh International Conference on Advanced Electronic Materials, Computers, and Software Engineering (AEMCSE 2024); 132292L (2024) https://doi.org/10.1117/12.3037941
Event: Seventh International Conference on Advanced Electronic Materials, Computers, and Software Engineering (AEMCSE 2024), 2024, Nanchang, China

Abstract

Recently, deep generative models have made significant progress in handling missing data, with the most notable being the use of variational autoencoders (VAEs). However, most of the VAEs methods project the original data into a lowdimensional vector space by the deep latent models. This can cause the loss of spatial information for matrix data, such as images, resulting in biased imputation. To address this issue, we propose an unsupervised learning method based on matrix-variate variational autoencoder, which generates a low-dimensional latent matrix through a convolutional network and reconstruct it through deconvolution. Besides, we proposed an iterative imputation method for better image reconstruction. Through testing on classic image datasets, the experimental results effectiveness of the proposed method.

(2024) Published by SPIE. Downloading of the abstract is permitted for personal use only.

Citation Download Citation

Shenfen Kuang "Unsupervised data imputation via matrix-variate variational autoencoders", Proc. SPIE 13229, Seventh International Conference on Advanced Electronic Materials, Computers, and Software Engineering (AEMCSE 2024), 132292L (7 August 2024); https://doi.org/10.1117/12.3037941

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available