Latent Nonlinear Denoising Score Matching for Enhanced Learning of Structured Distributions
This work addresses a specific bottleneck in generative modeling for structured data like MNIST, representing an incremental improvement over existing latent SGM frameworks.
The paper tackles the problem of slow synthesis and poor learning of structured distributions in score-based generative models by introducing latent nonlinear denoising score matching (LNDSM), which achieves faster synthesis and superior sample quality and variability compared to benchmark methods.
We present latent nonlinear denoising score matching (LNDSM), a novel training objective for score-based generative models that integrates nonlinear forward dynamics with the VAE-based latent SGM framework. This combination is achieved by reformulating the cross-entropy term using the approximate Gaussian transition induced by the Euler-Maruyama scheme. To ensure numerical stability, we identify and remove two zero-mean but variance exploding terms arising from small time steps. Experiments on variants of the MNIST dataset demonstrate that the proposed method achieves faster synthesis and enhanced learning of inherently structured distributions. Compared to benchmark structure-agnostic latent SGMs, LNDSM consistently attains superior sample quality and variability.