Timezone: »

Laplacian Autoencoders for Learning Stochastic Representations
Marco Miani · Frederik Warburg · Pablo Moreno-Muñoz · Nicki Skafte · Søren Hauberg

Tue Nov 29 09:00 AM -- 11:00 AM (PST) @ Hall J #428

Established methods for unsupervised representation learning such as variational autoencoders produce none or poorly calibrated uncertainty estimates making it difficult to evaluate if learned representations are stable and reliable. In this work, we present a Bayesian autoencoder for unsupervised representation learning, which is trained using a novel variational lower-bound of the autoencoder evidence. This is maximized using Monte Carlo EM with a variational distribution that takes the shape of a Laplace approximation. We develop a new Hessian approximation that scales linearly with data size allowing us to model high-dimensional data. Empirically, we show that our Laplacian autoencoder estimates well-calibrated uncertainties in both latent and output space. We demonstrate that this results in improved performance across a multitude of downstream tasks.

Author Information

Marco Miani (Technical University of Denmark)
Frederik Warburg (Technical University of Denmark)
Pablo Moreno-Muñoz (Technical University of Denmark (DTU))
Nicki Skafte (Technical University of Denmark)
Søren Hauberg (Technical University of Denmark)

More from the Same Authors