• Corpus ID: 227274821

The Autoencoding Variational Autoencoder

  title={The Autoencoding Variational Autoencoder},
  author={Ali Taylan Cemgil and Sumedh Ghaisas and Krishnamurthy Dvijotham and Sven Gowal and Pushmeet Kohli},
Does a Variational AutoEncoder (VAE) consistently encode typical samples generated from its decoder? This paper shows that the perhaps surprising answer to this question is `No'; a (nominally trained) VAE does not necessarily amortize inference for typical samples that it is capable of generating. We study the implications of this behaviour on the learned representations and also the consequences of fixing it by introducing a notion of self consistency. Our approach hinges on an alternative… 

Figures and Tables from this paper

Defending Variational Autoencoders from Adversarial Attacks with MCMC
This work examines several objective functions for adversarial attacks construction, suggest metrics assess the model robustness, and propose a solution to alleviate the effect of an attack, which utilizes the Markov Chain Monte Carlo technique in the inference step and is motivated by the theoretical analysis.
Certifiably Robust Variational Autoencoders
An approach for training variational autoencoders (VAEs) that are certifiably robust to adversarial attack by derive actionable bounds on the minimal size of an input perturbation re-quired to change a VAE’s reconstruction by more than an allowed amount and show how these parameters can be controlled.
Exploring the Latent Space of Autoencoders with Interventional Assays
A framework, called latent responses, is proposed, which exploits the locally contractive behavior exhibited by variational autoencoders to explore the learned manifold and extends the notion of disentanglement to take the learned generative process into account and consequently avoid the limitations of existing metrics that may rely on spurious correlations.
Generalization Gap in Amortized Inference
This work proposes a new training objective, inspired by the classic wake-sleep algorithm, to improve the generalizations properties of amortized inference and demonstrates how it can improve generalization performance in the context of image modeling and lossless compression.
Interventional Assays for the Latent Space of Autoencoders
This work investigates "holes" in the representation to quantitatively ascertain to what extent the latent space of a trained VAE is consistent with the chosen prior, and uses the identified structure to improve interpolation between latent vectors.
Boosting Semi-Supervised Anomaly Detection via Contrasting Synthetic Images
  • Sheng-Feng Yu, W. Chiu
  • Computer Science
    2021 17th International Conference on Machine Vision and Applications (MVA)
  • 2021
This paper proposes a simple but effective extension to further boost the performance of anomaly detection via integrating the contrastive learning and the generative model of inliers into a unified framework.
Lifelong Generative Modelling Using Dynamic Expansion Graph Model
  • Fei Ye, A. Bors
  • Computer Science
    Proceedings of the AAAI Conference on Artificial Intelligence
  • 2022
This paper studies the forgetting behaviour of VAEs using a joint GR and ENA methodology, by deriving an upper bound on the negative marginal log-likelihood, and proposes a novel Dynamic Expansion Graph Model (DEGM), which guarantees optimal performance for each task while also minimizing the required number of parameters.


Adversarially Robust Representations with Smooth Encoders
This paper proposes to augment the data with specifications that enforce insensitivity of the representation with respect to families of transformations, and proposes a regularization method based on a selection mechanism that creates a fictive data point by explicitly perturbing an observed true data point.
What regularized auto-encoders learn from the data-generating distribution
It is shown that the auto-encoder captures the score (derivative of the log-density with respect to the input) and contradicts previous interpretations of reconstruction error as an energy function.
InfoVAE: Balancing Learning and Inference in Variational Autoencoders
It is shown that the proposed Info-VAE model can significantly improve the quality of the variational posterior and can make effective use of the latent features regardless of the flexibility of the decoding distribution.
Recent Advances in Autoencoder-Based Representation Learning
An in-depth review of recent advances in representation learning with a focus on autoencoder-based models and makes use of meta-priors believed useful for downstream tasks, such as disentanglement and hierarchical organization of features.
Disentangling Factors of Variation with Cycle-Consistent Variational Auto-Encoders
This paper introduces a novel architecture that disentangles the latent space into two complementary subspaces by using only weak supervision in form of pairwise similarity labels in a variational auto-encoder framework.
Correlated Variational Auto-Encoders
An approximation by average of a set of tractable lower bounds over all maximal acyclic subgraphs of the undirected correlation graph is developed to address the intractability introduced by the correlated prior.
GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium
This work proposes a two time-scale update rule (TTUR) for training GANs with stochastic gradient descent on arbitrary GAN loss functions and introduces the "Frechet Inception Distance" (FID) which captures the similarity of generated images to real ones better than the Inception Score.
Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representations
This paper theoretically shows that the unsupervised learning of disentangled representations is fundamentally impossible without inductive biases on both the models and the data, and trains more than 12000 models covering most prominent methods and evaluation metrics on seven different data sets.
Understanding disentangling in β-VAE
A modification to the training regime of β-VAE is proposed, that progressively increases the information capacity of the latent code during training, to facilitate the robust learning of disentangled representations in β- VAE, without the previous trade-off in reconstruction accuracy.
Emergence of Invariance and Disentanglement in Deep Representations
It is shown that in a deep neural network invariance to nuisance factors is equivalent to information minimality of the learned representation, and that stacking layers and injecting noise during training naturally bias the network towards learning invariant representations.