A model must adapt itself to generalize to new and different data during testing. In this setting of fully test-time adaptation the model has only the test data and its own parameters. We propose to adapt by test entropy minimization (tent): we optimize the model for confidence as measured by the entropy of its predictions. Tent reduces generalization error for image classification on corrupted...

The reduction of mechanical dissipation increases force sensitivity of mechanical oscillators and decreases the thermal decoherence rate of mechanical quantum states. Concomitantly, reduced size allows stronger coupling to other degrees of freedom such as electromagnetic fields. These two demands are conflicting: due to surface losses, smaller size typically leads to increased dissipation....

From potential contamination of individuals with radioactive fission products after a nuclear accident to the therapeutic use of radio-isotopes for cancer diagnostics and treatment, the coordination and biological chemistry of actinides have become increasingly relevant to a number of applied problems. Understanding the fundamental bonding interactions of selective metal assemblies presents a...

In this talk, I will discuss advances in photonic engineering in which algorithmic approaches to device implementation unlock new functional capabilities for wavefront engineering. In the first part, I will discuss the utilization of freeform optimization for metamaterials in which non-local, multiple scattering dynamics enable new regimes of device efficiency and multi-functionality. These...

The production of polymer products relies largely on age-old molding techniques. A major reason for this is that additive methods have not delivered meaningful alternatives to traditional processes—until now. In this talk, I will describe Continuous Liquid Interface Production (CLIP) technology, which embodies a convergence of advances in software, hardware, and materials to bring the digital...