Adversarial examples trained on an ensemble of CNNs with a retinal preprocessing layer reduce the accuracy of time-limited humans in a two alternative forced choice task.
The Control What You Can (CWYC) method learns to control components of the environment to achieve multi-step goals by combining task planning with surprise and learning progress based intrinsic motivation.
Introduces a novel hierarchical representation of visual scenes, Physical Scene Graphs (PSGs), as well as a network for learning them from RGB movies, PSGNet, which outperforms other unsupervised methods in scene segmentation.
Novel method for video prediction from a single frame by decomposing the scene into entities with location and appearance features, capturing ambiguities with a global latent variable.
Proposes multitask model to jointly learn semantic goal navigation and embodied question answering.
Investigates automatically generating curricula based on a variety of progress signals that are computed for each data sample.
Applies curriculum learning to a RL context to achieve policies.