Beyond existing meta-RL analyses, we provide upper bounds of the expected optimality gap over the task distribution.

However,existing methods lackrobustness as they may decide to explore areas of the latent space for which no data was available during training andwhere thedecoder canbeunreliable, leading tothe generation ofunrealistic orinvalidobjects. Wepropose toleveragetheepistemic uncertainty of the decoder to guide the optimization process.

In this work, we focus on the task of controlling the level of sparsity when performing sparse learning.

In this work, we focus on the task of controlling the level of sparsity when performing sparse learning.