DeepGreen: deep learning of Green's functions for nonlinear boundary value problems.

Boundary value problems (BVPs) play a central role in the mathematical analysis of constrained physical systems subjected to external forces. Consequently, BVPs frequently emerge in nearly every engineering discipline and span problem domains including fluid mechanics, electromagnetics, quantum mechanics, and elasticity. The fundamental solution, or Green's function, is a leading method for solving linear BVPs that enables facile computation of new solutions to systems under any external forcing. However, fundamental Green's function solutions for nonlinear BVPs are not feasible since linear superposition no longer holds. In this work, we propose a flexible deep learning approach to solve nonlinear BVPs using a dual-autoencoder architecture. The autoencoders discover an invertible coordinate transform that linearizes the nonlinear BVP and identifies both a linear operator L and Green's function G which can be used to solve new nonlinear BVPs. We find that the method succeeds on a variety of nonlinear systems including nonlinear Helmholtz and Sturm-Liouville problems, nonlinear elasticity, and a 2D nonlinear Poisson equation and can solve nonlinear BVPs at orders of magnitude faster than traditional methods without the need for an initial guess. The method merges the strengths of the universal approximation capabilities of deep learning with the physics knowledge of Green's functions to yield a flexible tool for identifying fundamental solutions to a variety of nonlinear systems.

The reliability and predictive validity of the Guidelines for Stalking Assessment and Management (SAM).

This study assessed the reliability and validity of the Guidelines for Stalking Assessment and Management (SAM), a structured professional judgment measure for assessing stalking risks. The SAM was completed retrospectively from file review for 146 adult stalkers (90.4% male) referred to a community-based forensic mental health service. Interrater reliability (IRR) was initially poor, but developing a strict definition of stalking currency and rescoring the SAM led to improvement. Based on the updated scoring, IRR was moderate for judgments about whether stalking was ongoing at the time of assessment, and fair to moderate for summary risk judgments. Both case prioritization (area under the curve [AUC] = .69) and risk for continued stalking (AUC = .76) ratings discriminated between groups, with high-risk stalkers 5-9 times as likely as low-risk stalkers to reoffend by stalking their original victims. Lifetime SAM total scores (AUC = .70) also featured moderate to good discrimination. Follow-up analyses suggested that this was driven mainly by the recent presence of risk markers and the nature of any ongoing stalking situation rather than historical or individual factors. Findings support the use of the SAM to structure risk judgments made by those with experience in assessing stalking. Current results also imply that IRR might be improved by introducing (a) a fixed definition of stalking currency and (b) usage guidelines for specific contexts. (PsycINFO Database Record (c) 2018 APA, all rights reserved).

The Reliability and Predictive Validity of the Stalking Risk Profile.

This study assessed the reliability and validity of the Stalking Risk Profile (SRP), a structured measure for assessing stalking risks. The SRP was administered at the point of assessment or retrospectively from file review for 241 adult stalkers (91% male) referred to a community-based forensic mental health service. Interrater reliability was high for stalker type, and moderate-to-substantial for risk judgments and domain scores. Evidence for predictive validity and discrimination between stalking recidivists and nonrecidivists for risk judgments depended on follow-up duration. Discrimination was moderate (area under the curve = 0.66-0.68) and positive and negative predictive values good over the full follow-up period ( Mdn = 170.43 weeks). At 6 months, discrimination was better than chance only for judgments related to stalking of new victims (area under the curve = 0.75); however, high-risk stalkers still reoffended against their original victim(s) 2 to 4 times as often as low-risk stalkers. Implications for the clinical utility and refinement of the SRP are discussed.

Microfabrication of human organs-on-chips.

'Organs-on-chips' are microengineered biomimetic systems containing microfluidic channels lined by living human cells, which replicate key functional units of living organs to reconstitute integrated human organ-level pathophysiology in vitro. These microdevices can be used to test efficacy and toxicity of drugs and chemicals, and to create in vitro models of human disease. Thus, they potentially represent low-cost alternatives to conventional animal models for pharmaceutical, chemical and environmental applications. Here we describe a protocol for the fabrication, microengineering and operation of these microfluidic organ-on-chip systems. First, microengineering is used to fabricate a multilayered microfluidic device that contains two parallel elastomeric microchannels separated by a thin porous flexible membrane, along with two full-height, hollow vacuum chambers on either side; this requires ∼3.5 d to complete. To create a 'breathing' lung-on-a-chip that mimics the mechanically active alveolar-capillary interface of the living human lung, human alveolar epithelial cells and microvascular endothelial cells are cultured in the microdevice with physiological flow and cyclic suction applied to the side chambers to reproduce rhythmic breathing movements. We describe how this protocol can be easily adapted to develop other human organ chips, such as a gut-on-a-chip lined by human intestinal epithelial cells that experiences peristalsis-like motions and trickling fluid flow. Also, we discuss experimental techniques that can be used to analyze the cells in these organ-on-chip devices.