#generative modeling
6 resultsImage Synthesis with a Single (Robust) Classifier
Shibani Santurkar, Dimitris Tsipras, Brandon Tran +3
The paper demonstrates that a single off‑the‑shelf classifier, when trained to be adversarially robust, can be used for various image synthesis tasks by directly manipulating salie…
Audio Source Separation Using Variational Autoencoders and Weak Class Supervision
ErtuÄ Karamatlı, Ali Taylan Cemgil, Serap Kırbız
The paper introduces a method for separating audio sources in mixtures using class‑level labels only, by attaching a variational autoencoder to each source class within a non‑negat…
Noise Contrastive Variational Autoencoders
Octavian-Eugen Ganea, Yashas Annadani, Gary Bécigneul
The paper studies why variational autoencoders often suffer from posterior collapse and proposes a new model, NC‑VAE, that uses noise contrastive estimation to prevent this collaps…
Training capsules as a routing-weighted product of expert neurons
Michael Hauser
The paper models capsule networks as a product of expert neurons, using routing‑by‑agreement coefficients as weights in an energy‑based formulation, and introduces an unsupervised…
Training products of expert capsules with mixing by dynamic routing
Michael Hauser
The paper proposes an unsupervised learning algorithm for capsule networks that uses an energy function and dynamic routing to train products of expert capsules, enabling realistic…
Information processing constraints in travel behaviour modelling: A generative learning approach
Melvin Wong, Bilal Farooq
The paper introduces a data‑driven generative model based on rational inattention theory to capture how travelers process limited information and uncertainty when making travel cho…