One humid afternoon in July, José Ramírez-Garofalo drove his large Toyota truck through the lush new hills, valleys, and meadows of Freshkills Park, a twenty-two-hundred-acre green space that the city is constructing on Staten Island. Ramírez-Garofalo, a young man with dark hair, large forearms, and the beginnings of a goatee, drove and talked fast. "It's an impermeable geotextile membrane," he said, referring to the thick plastic that was used, starting in the mid-nineties, to cap the four giant trash mounds of the old Fresh Kills landfill. "On top there is playground soil." The process of capping and terraforming the four mounds that once made up the country's largest dump is complete, but the park won't be fully open until at least 2036.

On a number of common benchmark tasks to recover underlying expressions from a dataset, our method recovers 65% more expressions than a recently published top-performing model using the same experimental setup.

The vital part of every learning-based algorithm is an optimization procedure, e.g., Stochastic Gradient Descent.

In this context, we present a teacher-student framework that can distill the black-box (teacher) model into a student model with minimal accuracy loss.

We thank reviewers for their useful comments and insights. We next address concerns raised by the reviewers. The evolutionary algorithm might collapse to certain regions of the output space. This demonstrates that the proposed evolutionary strategy does not collapse to certain regions. Did the authors look at mode collapse of GAN or blurry effect of V AE?