Training VAEs Under Structured Residuals
نویسندگان
چکیده
Variational auto-encoders (VAEs) are a popular and powerful deep generative model. Previous works on VAEs have assumed a factorised likelihood model, whereby the output uncertainty of each pixel is assumed to be independent. This approximation is clearly limited as demonstrated by observing a residual image from a VAE reconstruction, which often possess a high level of structure. This paper demonstrates a novel scheme to incorporate a structured Gaussian likelihood prediction network within the VAE that allows the residual correlations to be modelled. Our novel architecture, with minimal increase in complexity, incorporates the covariance matrix prediction within the VAE. We also propose a new mechanism for allowing structured uncertainty on color images. Furthermore, we provide a scheme for effectively training this model, and include some suggestions for improving performance in terms of efficiency or modelling longer range correlations. The advantage of our approach is illustrated on the CelebA face data and the LSUN outdoor churches dataset, with substantial improvements in terms of samples over traditional VAE and better reconstructions.
منابع مشابه
Observer-based fault detection and isolation for structured systems
Fault Detection and Isolation (FDI) problems are here considered for linear systems with faults and disturbances. We design a set of observer-based residuals, in such a way that the transfer from the disturbances to the residuals is zero and the transfer from the faults to the residuals allows fault isolation. We are interested in obtaining a transfer function from faults to residuals with eith...
متن کاملAdversarial Variational Bayes: Unifying Variational Autoencoders and Generative Adversarial Networks
Variational Autoencoders (VAEs) are expressive latent variable models that can be used to learn complex probability distributions from training data. However, the quality of the resulting model crucially relies on the expressiveness of the inference model. We introduce Adversarial Variational Bayes (AVB), a technique for training Variational Autoencoders with arbitrarily expressive inference mo...
متن کاملL Earning to I Nfer
Inference models, which replace an optimization-based inference procedure with a learned model, have been fundamental in advancing Bayesian deep learning, the most notable example being variational auto-encoders (VAEs). In this paper, we propose iterative inference models, which learn how to optimize a variational lower bound through repeatedly encoding gradients. Our approach generalizes VAEs ...
متن کاملInducing Interpretable Representations with Variational Autoencoders
We develop a framework for incorporating structured graphical models in the encoders of variational autoencoders (VAEs) that allows us to induce interpretable representations through approximate variational inference. This allows us to both perform reasoning (e.g. classification) under the structural constraints of a given graphical model, and use deep generative models to deal with messy, high...
متن کاملStructured Fault Diagnosis in Mildly Nonlinear Systems : Parity Space and Input-output Formulations
Concerned is residual generation in systems with mild nonlinearities, that is, nonlinearities which involve the inputs, outputs and faults but not the state. It is shown how structured residuals can be designed by the parity space method, in discrete time, if mild nonlinearities are present. Next a number of fundamental links between the parity space and the input-output representations are exp...
متن کامل