We present a novel method for simultaneous learning of depth, egomotion, object motion, and camera intrinsics from monocular videos, using only consistency across neighboring video frames as supervision signal. Similarly to prior work, our method learns by applying differentiable warping to frames and comparing the result to adjacent ones, but it provides several improvements: We address occlusions geometrically and differentiably, directly using the depth maps as predicted during training. We introduce randomized layer normalization, a novel powerful regularizer, and we account for object motion relative to the scene. To the best of our knowledge, our work is the first to learn the camera intrinsic parameters, including lens distortion, from video in an unsupervised manner, thereby allowing us to extract accurate depth and motion from arbitrary videos of unknown origin at scale. We evaluate our results on the Cityscapes, KITTI and Eu-RoC datasets, establishing new state of the art on depth prediction and odometry, and demonstrate qualitatively that depth prediction can be learned from a collection of YouTube videos.

The actin-binding protein filamin links membrane receptors to the underlying cytoskeleton. The cytoplasmic domains of these membrane receptors have been shown to bind to various filamin immunoglobulin repeats. Notably, among 24 human filamin repeats, repeat 17 was reported to specifically bind to platelet receptor glycoprotein Ib␣ and repeat 21 to integrins. However, a complete sequence alignment of all 24 human filamin repeats reveals that repeats 17 and 21 actually belong to a distinct filamin repeat subgroup (containing repeats 4, 9, 12, 17, 19, 21, and 23) that shares a conserved ligand-binding site. Using isothermal calorimetry and NMR analyses, we show that all repeats in this subgroup can actually bind glycoprotein Ib␣, integrins, and a cytoskeleton regulator migfilin in similar manners. These data provide a new view on the ligand specificity of the filamin repeats. They also suggest a multiple ligand binding mechanism where similar repeats within a filamin monomer may promote receptor clustering or receptor cross-talking for regulation of the cytoskeleton organization and diverse filaminmediated cellular activities.

The management options for ureteral obstruction are diverse, including retrograde ureteral stent insertion or antegrade nephrostomy placement, with or without eventual antegrade stent insertion. There is currently no consensus on the ideal treatment or treatment pathway for ureteral obstruction owing, in part, to the varied etiologies of obstruction and diversity of institutional practices. Additionally, different clinicians such as internists, urologists, oncologists and radiologists are often involved in the care of patients with ureteral obstruction and may have differing opinions concerning the best management strategy. The purpose of this manuscript was to review available literature that compares percutaneous nephrostomy placement vs ureteral stenting in the management of ureteral obstruction from both benign and malignant etiologies.

We consider the task of deterministically entangling two remote qubits using joint measurement and feedback, but no directly entangling Hamiltonian. In order to formulate the most effective experimentally feasible protocol, we introduce the notion of average sense locally optimal (ASLO) feedback protocols, which do not require real-time quantum state estimation, a difficult component of real-time quantum feedback control. We use this notion of optimality to construct two protocols which can deterministically create maximal entanglement: a semiclassical feedback protocol for low efficiency measurements and a quantum feedback protocol for high efficiency measurements. The latter reduces to direct feedback in the continuous-time limit, whose dynamics can be modeled by a Wiseman-Milburn feedback master equation which yields an analytic solution in the limit of unit measurement efficiency. Our formalism can smoothly interpolate between continuous-time and discrete-time descriptions of feedback dynamics, and we exploit this feature to then derive a superior hybrid protocol for arbitrary non-unit measurement efficiency that switches between quantum and semiclassical protocols. Finally, we show using simulations incorporating experimental imperfections that deterministic entanglement of remote superconducting qubits may be achieved with current technology using the continuous-time feedback protocol alone.

Because of clerical errors in preparing the figures, in Fig. 1A, the last portions of the mouse and human TILRR sequences are not aligned with the consensus sequence, and the human form is mislabeled as 716 amino acids (aa). In Fig. 2C (and on page 7227, right column, line 4), the most potent form of the human protein is mislabeled as 710 aa. Supplemental Fig. S1 correctly shows the alignment and the length of the human TILRR protein as 715 aa, with the most potent form, lacking the N-terminal 6 aa, as 709 aa. The amino acid sequence is correct as shown in all figures, and the clerical errors have no impact on any of the results, including the function of the protein, the probes used, or the numbering of the mutants.