University of Southern California

Neuromorphic computing and artificial neurons have been shown to improve the solution for some of the complex problems for conventional computers. We present a spiking soma (JJ-Soma) circuit that consists of a double-junction SQUID interfered with a resistor (threshold loop), a decaying superconductor loop cut by a resistor, which is coupled to the SQUID like structure. The proposed soma has three main properties: (i) ultra-high-speed operation with minimal power consumption, (ii) compatibility with standard foundry processes that allows fabrication with the available infrastructure, (iii)and compatibility with conventional SFQ logic gates which enables design and implementation of complicated networks. Each soma circuit covers 40 μm × 80 μm area on-chip with different activation functions. The circuits are fabricated in a commercial foundry for superconductors and they have been implemented and demonstrated experimentally.

In five and a half years, the ColdFlux project under the IARPA SuperTools program pushed the boundaries of understanding the requirements for and the actual capabilities of digital and analog superconductor electronic design automation (S-EDA) tools. The SuperTools program demanded significant beyond-state-of-the-art deliverables in four main areas: RTL synthesis, architectures, and verification; analog design and layout synthesis; physical design and test; and technology CAD and cell library design. Through the work of academic groups scattered over four continents, the ColdFlux effort forged into a powerful set of open-source and commercial S-EDA tools unlike any before, rivaled only by a commercial toolchain from Synopsys under the same SuperTools umbrella. We present an overview of the tools from where we started to the eventual project deliverables before showcasing some of the most significant progress–catalyzed by the project's demands–beyond the original program goals. These include powerful and fast simulation and extraction engines, magnetic field and flux trapping analysis, advanced clocking methods, multi-chip interface extraction and verification, superconductor multi-port S-parameter extraction, unified multi-layer design-rule compliant track blocks for automated place and route of both rapid single flux quantum (RSFQ) and adiabatic quantum-flux-parametron (AQFP) cells, models and tools for validation and test, multi-bit single flux quantum (SFQ) cell design and their usage in architecture innovations for full CPU designs and more. Comprehensive RSFQ and AQFP cell libraries and a process design kit (PDK) were developed with the ColdFlux tools as part of the project. The AQFP cell library features a logically rich collection of 80+ cells, including 3- and 5-input logic gates, signal-driving boosters, and refined RSFQ-to-AQFP interfaces, while the RSFQ library has 30+ cells. We also discuss the efforts to make the tools work over multiple operating systems, maintain and disseminate open-source code, and simplify compilation and installation. We include links to all the code repositories. Finally, we discuss how the full toolchain enables and enhances the superconductor IC design process.

This study investigated how corporate social responsibility, adherence to social missions, organisation–public relationships outcomes, and corporate reputation influence the purchase intentions for social enterprises’ products. Through a survey of Taiwanese consumers (N = 507) in online channels, results suggested that corporate social responsibility and adherence to social missions indirectly influenced purchase intentions through organisation−public relationships and corporate reputation. However, the direct effect of corporate social responsibility on organisation–public relationship outcomes was nonsignificant. Adherence to social missions acted as a more important factor than corporate social responsibility in generating purchase intentions, indicating the importance of social missions in the social enterprise context.

To ensure the performance of large-scale datacenters, operators need to monitor up to tens of millions of various-type KPIs, e.g., CPU utilization, memory utilization. For each KPI, it is crucial but challenging to detect outliers that deviate from its historical patterns or the patterns of other KPIs in the same period. In this work, we propose OutSpot , an unsupervised outlier detection framework that integrates hierarchical agglomerative clustering (HAC) with conditional variational autoencoder (CVAE), which significantly improves computational efficiency and comprehensively learns the above two patterns. Additionally, two simple yet effective techniques, soft threshold and median filter, are applied to precisely determine outlier KPIs. Using two real-world datasets collected from the datacenters owned by a top-tier global short video service provider and a top-tier domestic operator,respectively. It demonstrates that OutSpot achieves the best F1 score of 0.95 and 0.91, AUC of 0.99 and 0.99 on the two datasets, significantly outperforming seven baseline outlier detection methods.

Artificial neurons provide a new way of computation for neuro-inspired algorithms, and the abilities may efficiently solve the challenges. We propose implementations of logic gates (AND, OR, XOR, Majority), full adder, full subtractor, even parity generator, and 2-bit multiplier circuit formed by JJ-Soma and standard RSFQ digital library cells. The designed circuits execute ultra-high-speed operations without a clock signal, and they are capable of processing parallel or time-sliced operations. The combination of JJ-Soma cells and SFQ cells creates the potential models for the arithmetic logic unit devices with a small on-chip area, high operating speed, and pipeline structures for microprocessors. In this study, the previously optimized JJ-Soma circuits have low power consumption and high computational speed where the firing rates for two pulse and three pulse threshold circuits were designed to be 50 and 15 GHz with about 10 $^{-19}$ J/spike energy level. The proposed circuits, fabricated with a commercial foundry service, have been implemented and demonstrated experimentally.

Pre-training a diverse set of neural network controllers in simulation has enabled robots to adapt online to damage in robot locomotion tasks. However, finding diverse, high-performing controllers requires expensive network training and extensive tuning of a large number of hyperparameters. On the other hand, Covariance Matrix Adaptation MAP-Annealing (CMA-MAE), an evolution strategies (ES)-based quality diversity algorithm, does not have these limitations and has achieved state-of-the-art performance on standard QD benchmarks. However, CMA-MAE cannot scale to modern neural network controllers due to its quadratic complexity. We leverage efficient approximation methods in ES to propose three new CMA-MAE variants that scale to high dimensions. Our experiments show that the variants outperform ES-based baselines in benchmark robotic locomotion tasks, while being comparable with or exceeding state-of-the-art deep reinforcement learning-based quality diversity algorithms.

Cervical spinal pain, defined as pain perceived anywhere in the posterior region of the cervical spine and/or shoulder regions, is a common musculoskeletal disorder and a leading cause of disability worldwide. Approximately half of all individuals will experience at least one episode of clinically important neck pain over the course of their life. Cervical spinal pain can be caused by numerous etiologies and is commonly caused by a multitude of acute and chronic anatomic, behavioral, and psychosocial factors. Though symptoms resolve spontaneously in 50% of patients presenting with acute cervical spinal pain, the majority of patients continue to experience symptoms and seek treatment for years after their initial presentation. Therefore, diagnosing and classifying cervical spinal pain correctly are pertinent and proper workup of this potentially debilitating pathology is vital to direct efficacious treatment. Our chapter aims to provide a snapshot into cervical spinal pain, its various classifications, and effective means of diagnosing and treating neck pain.

Hepatic oncologic surgical emergencies can be approached using many core surgical principles familiar to general and acute care surgeons. In this chapter, surgical emergencies related to underlying liver oncology are discussed in detail by first reviewing the hepatic anatomy and physiology relevant to these emergencies. Next, an approach to emergent liver surgery is presented and centered upon exposures and concepts encountered in trauma surgery. Lastly, specific tumor- and treatment-related complications which may result in surgical emergencies among patients with hepatic malignancy are examined.

This study evaluates the beneficial effects of discharging nanosecond pulse transient plasma (NPTP) in a coaxial electrostatic precipitator for capturing nanoscale soot particles (∼50 nm) produced by an ethylene flame. Here, the nanoscale soot particles are collected using two different reactor geometries: a 3′′ diameter reactor with a mean flow velocity of 1.2 m s⁻¹ and a 1.5′′ diameter reactor with a mean flow velocity 1.5 m s⁻¹, corresponding to volumetric flow rates of 11.5 CFM and 3.6 CFM, respectively. The nanosecond high voltage pulses (+20 kV, 20 ns, 800 Hz) are applied in conjunction with DC bias voltages. While nearly 100% collection efficiency can be achieved without NPTP at sufficiently high DC voltages (|VDC| > 14 kV), this drops below 50% for lower DC voltages (|VDC| < 10 kV). With NPTP, we observe substantially enhanced remediation (up to 23×) at lower DC voltages (|VDC| < 10 kV) due to the enhanced ion density produced by the plasma. For DC-only electrostatic precipitation, the charging of soot particles takes place via a DC corona, whose ion density is several orders of magnitude lower than that of the NPTP, which produces a streamer discharge due to the fast rise times of the nanosecond pulses (i.e., dV/dt ∼ 10¹² V s⁻¹). High speed imaging of the plasma emission profile indicates that ion densities 10⁶ times higher are achieved with the nanosecond pulsed plasma, as compared to that of the DC corona. At lower DC voltages (i.e., |VDC| < 10 kV), the charging of soot particles is a key factor limiting the DC-only remediation efficiencies, and NPTP provides a way to mitigate this limitation.

Objective Substantial practice variation exists in the management of children with nonsevere traumatic intracranial hemorrhage (tICH). A comprehensive understanding of rates and timing of clinically important tICH, including critical interventions and deterioration, along with associated clinical and neuroradiographic characteristics, will inform accurate risk stratification. Methods We conducted a single-center retrospective cohort study of children aged younger than 18 years evaluated in the emergency department (ED) from May 1, 2014 to February 28, 2020 with tICH and initial Glasgow Coma Scale (GCS) score of higher than 8. We determined rates of clinically important tICH after injury and within 96 hours of ED arrival, defined as immediate ED interventions (intubation, hyperosmotic agents, or neurosurgery within 4 hours of arrival) or clinically important deterioration (signs/symptoms with change in management). Associations between outcome and clinical and neuroradiographic characteristics were calculated using individual logistic regression models. Results Our sample included 135 children. Clinically important tICH was observed in 13.3% (n = 18); 9 (6.7%) underwent immediate ED interventions and 9 (6.7%) developed deterioration. Most (93.3%, n = 127) presented with an initial GCS ≥ 14, including all children who later deteriorated. Initial GCS ( P = 0.001) and nonaccidental trauma ( P = 0.024) mechanism were associated with the outcome. None of the 71 (52.6%) children with initial GCS ≥ 14, isolated, nonepidural hemorrhage after accidental injury developed clinically important tICH. Conclusions Clinically important tICH occurred in 13% of children with nonsevere tICH, and 7% of children who did not undergo immediate ED interventions later deteriorated, all of whom had an initial GCS ≥ 14. However, a subgroup of children was identified as low risk based on clinical and neuroradiographic characteristics.

Digital Twin (DT) is a novel concept that may bring a paradigm shift for precision medicine. In this study we demonstrate a DT application for estimating the age of onset of disease-specific brain atrophy in individuals with multiple sclerosis (MS) using brain MRI. We first augmented longitudinal data from a well-fitted spline model derived from a large cross-sectional normal aging data. Then we compared different mixed spline models through both simulated and real-life data and identified the mixed spline model with the best fit. Using the appropriate covariate structure selected from 52 different candidate structures, we augmented the thalamic atrophy trajectory over the lifespan for each individual MS patient and a corresponding hypothetical twin with normal aging. Theoretically, the age at which the brain atrophy trajectory of an MS patient deviates from the trajectory of their hypothetical healthy twin can be considered as the onset of progressive brain tissue loss. With a tenfold cross validation procedure through 1000 bootstrapping samples, we found the onset age of progressive brain tissue loss was, on average, 5–6 years prior to clinical symptom onset. Our novel approach also discovered two clear patterns of patient clusters: earlier onset versus simultaneous onset of brain atrophy.

Importance Although the barriers to dementia care in primary care are well characterized, primary care practitioner (PCP) perspectives could be used to support the design of values-aligned dementia care pathways that strengthen the role of primary care. Objective To describe PCP perspectives on their role in dementia diagnosis and care. Design, Setting, and Participation In this qualitative study, interviews were conducted with 39 PCPs (medical doctors, nurse practitioners, and doctors of osteopathic medicine) in California between March 2020 and November 2022. Results were analyzed using thematic analysis. Main Outcomes and Measures Overarching themes associated with PCP roles in dementia care. Results Interviews were conducted with 39 PCPs (25 [64.1%] were female; 16 [41%] were Asian). The majority (36 PCPs [92.3%]) reported that more than half of their patients were insured via MediCal, the California Medicaid program serving low-income individuals. Six themes were identified that convey PCPs’ perspectives on their role in dementia care. These themes focused on (1) their role as first point of contact and in the diagnostic workup; (2) the importance of long-term, trusting relationships with patients; (3) the value of understanding patients’ life contexts; (4) their work to involve and educate families; (5) their activities around coordinating dementia care; and (6) how the care they want to provide may be limited by systems-level constraints. Conclusions and Relevance In this qualitative study of PCP perspectives on their role in dementia care, there was alignment between PCP perspectives about the core values of primary care and their work diagnosing and providing care for people living with dementia. The study also identified a mismatch between these values and the health systems infrastructure for dementia care in their practice environment.

The target article focuses on evidence from nonlinguistic faculties to defend the claim that cognition generally traffics in language-of-thought (LoT)-type representations. This focus creates needed space to discuss the mounting accumulation of nonclassical evidence for LoT, but it also misses relevant work in linguistics that directly offers a perspective on specific hypotheses about candidate LoT representations.

The standard of care for managing a patient with an implant is to identify the item and to assess the relative safety of scanning the patient. Because the 1.5 T MR system is the most prevalent scanner in the world and 3 T is the highest field strength in widespread use, implants typically have “MR Conditional” (i.e., an item with demonstrated safety in the MR environment within defined conditions) labeling at 1.5 and/or 3 T only. This presents challenges for a facility that has a scanner operating at a field strength below 1.5 T when encountering a patient with an implant, because scanning the patient is considered “off‐label.” In this case, the supervising physician is responsible for deciding whether to scan the patient based on the risks associated with the implant and the benefit of magnetic resonance imaging (MRI). For a passive implant, the MRI safety‐related concerns are static magnetic field interactions (i.e., force and torque) and radiofrequency (RF) field‐induced heating. The worldwide utilization of scanners operating below 1.5 T combined with the increasing incidence of patients with implants that need MRI creates circumstances that include patients potentially being subjected to unsafe imaging conditions or being denied access to MRI because physicians often lack the knowledge to perform an assessment of risk vs. benefit. Thus, physicians must have a complete understanding of the MRI‐related safety issues that impact passive implants when managing patients with these products on scanners operating below 1.5 T. This monograph provides an overview of the various clinical MR systems operating below 1.5 T and discusses the MRI‐related factors that influence safety for passive implants. Suggestions are provided for the management of patients with passive implants labeled MR Conditional at 1.5 and/or 3 T, referred to scanners operating below 1.5 T. The purpose of this information is to empower supervising physicians with the essential knowledge to perform MRI exams confidently and safely in patients with passive implants. Level of Evidence 1 Technical Efficacy Stage 3