The paper proposes two score matching objectives for partially missing data, proves a finite-sample bound for the truncated importance-weighted variant, and shows improved graphical model edge recovery on stock and yeast data.
Learning Deep Energy Models: Contrastive Divergence vs. Amortized MLE
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
abstract
We propose a number of new algorithms for learning deep energy models and demonstrate their properties. We show that our SteinCD performs well in term of test likelihood, while SteinGAN performs well in terms of generating realistic looking images. Our results suggest promising directions for learning better models by combining GAN-style methods with traditional energy-based learning.
fields
stat.ML 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Score Matching With Missing Data
The paper proposes two score matching objectives for partially missing data, proves a finite-sample bound for the truncated importance-weighted variant, and shows improved graphical model edge recovery on stock and yeast data.