Generative Modeling

Deep Generative Classifier with Short Run Inference

Deep generative classifier uses Short Run Markov Chain Monte Carlo inference, Langevin dynamics, and backpropagation through time to achieve similar classification accuracy as an analogous discriminative classifier, i.e., a convolutional neural network, while it has the advantages that it can generate data, it can learn unsupervised with additional unlabeled data, and it exhibits robustness to adversarial attacks due to the stochasticity of the Langevin equation and the top-down architecture of the underlying generator network.

Learning Multi-Layer Latent Variable Model via Variational Optimization of Short Run MCMC for Approximate Inference

Short-run MCMC residual network outperforms a variational autoencoder in terms of image reconstruction error and image synthesis quality, while not requiring the design of a separate inference network

Topology Adaptive Snakes Improve Mask Generation for Image Inpainting

T-snake deformable models, which can segment complex-shaped structures, improve generated masks passed to a GAN for image inpainting

Formulation of Variational Lower Bound and Application of VAE to MNIST Dataset

Formulation of the evidence lower bound (ELBO) for the variational autoencoder and an application to synthesizing binary images