.png){kind=icon}
Return
Tensor Density Estimator by Convolution-Deconvolution
.png){kind=icon}
PRE
AI
DOI：10.1016/j.acha.2026.101944.png)
.png){kind=icon}
.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
We propose a linear algebraic framework for performing density estimation. It consists of three simple steps: transform the empirical distribution into the coefficient domain and downweight some high-order coefficients to remove the exponentially large variance; compress the resulting high-dimensional coefficient tensor as a tensor train by proposed fast tensor decomposition algorithms; and finally, upweight the high-order coefficients and transform the coefficient tensor back to the original domain of the density. Numerical results demonstrate the high accuracy and efficiency of our proposed methods.
Keywords:
High-dimensional density estimation
tensor network
generative modeling
model compression
Journal
IF:
3.2
Papers:
127
Citations:
3.9K
Organization
Cited Papers
ShareSaveA sparse matrix arithmetic based on H-matrices.: Part I :: Introduction to H-matrices.png)
COMPUTING
COMPUTINGIF2.8
PREAI
ShareSaveShareSave
ShareSaveShareSaveShareSaveShareSave
View more