Explore the vast range of titles available.

Neuromorphic computation draws inspiration from the remarkable features of the human brain including low energy consumption, parallelism, adaptivity, cognitive functions, and learning ability. These qualities hold the promise of unlocking groundbreaking computational techniques that surpass the limitations of traditional computing systems. This paper reports a remarkable photo‐synaptic behavior in the field of rare earth ion‐doped luminescent oxides by using long‐persistent luminescence (LPL). This system utilizes electron trap states to regulate the synaptic behavior, operating through a fundamentally different mechanism from that of electronic‐based synaptic devices. To realize this strategy, Tb3+ doped CaSnO3, which shows a significant LPL property under UV‐light excitation, is prepared. The luminescent system shows key neuromorphic characteristics such as paired‐pulse facilitation, pulse‐number/timing dependent potentiation, and pulse‐number/timing dependent short‐ to long‐term plasticity transition, which are required for realizing synaptic devices. This feature expands the way for advanced neuromorphic technologies employing light stimuli. The principal synaptic features in biological synapses are successfully replicated in Tb3+ and Al3+ co‐doped CaSnO3 by utilizing long‐persistent luminescence. Upon the introduction of square‐wave UV pulses, this sample exhibits a sequential increase in photoluminescence intensity with both pulse number and frequency. These findings hold the potential for the development of photonic artificial synapses using luminescent oxides.

This study presents a design approach for a uniform porous surface to control laminar-to-turbulent transition in hypersonic boundary layers. The focus is on suppressing the Mack second mode, which is a dominant instability in hypersonic boundary layers. The Mack second mode is acoustic-wave-like in the ultrasonic frequency range and can be effectively attenuated by porous surfaces. Previous studies have explored porous surfaces, either by targeting a specific frequency or by adopting geometrically complex configurations for various frequencies. In contrast, the present study proposes a porous surface design that effectively stabilizes the Mack second mode over a wide frequency range, while maintaining structural simplicity. In addition, this porous surface design incorporates constraints associated with practical fabrication to enhance manufacturability. The absorption characteristics of porous surfaces are evaluated with an acoustic impedance model, and the stabilization performance is assessed with linear stability theory. The proposed porous surface design is compared with a conventional design method that focuses on the Mack second mode with a single frequency. Consequently, the proposed design methodology demonstrates robust and consistent suppression of the Mack second mode in a broad frequency range. This approach improves both stabilization performance and manufacturability with a uniform porous surface, contributing to its practical application in high-speed vehicles.

This paper proposes a novel circuit called the Line-Selector Unit (LSU) that enhances conventional oxide thin-film-transistor (TFT) scan drivers. The LSU, consisting of four transistors, one capacitor, and three additional control signals, allows a traditional scan driver to generate scan signals only in the selected area. A 12-stage scan driver with the LSU was fabricated using oxide TFTs, and experimental results demonstrate that the LSU operates regardless of the position and size of the selected area. Moreover, the LSU does not experience voltage stress during conventional scan driving and enhances reliability by employing an optimized driving method that distributes positive voltage bias stress during selective-scan driving. In addition, the LSU consumes only 0.1% of the total scan driver power consumption, making it a highly compatible and energy-efficient solution for low-power display applications.

Neuromorphic computation draws inspiration from the remarkable features of the human brain including low energy consumption, parallelism, adaptivity, cognitive functions, and learning ability. These qualities hold the promise of unlocking groundbreaking computational techniques that surpass the limitations of traditional computing systems. This paper reports a remarkable photo‐synaptic behavior in the field of rare earth ion‐doped luminescent oxides by using long‐persistent luminescence (LPL). This system utilizes electron trap states to regulate the synaptic behavior, operating through a fundamentally different mechanism from that of electronic‐based synaptic devices. To realize this strategy, Tb 3+ doped CaSnO 3 , which shows a significant LPL property under UV‐light excitation, is prepared. The luminescent system shows key neuromorphic characteristics such as paired‐pulse facilitation, pulse‐number/timing dependent potentiation, and pulse‐number/timing dependent short‐ to long‐term plasticity transition, which are required for realizing synaptic devices. This feature expands the way for advanced neuromorphic technologies employing light stimuli.

In this study, the gaseous nitriding was carried out for enhancing surface properties of AISI 1015 steel, and the formation of compound layer was investigated accordingly. For AISI 1015 steel, the nitriding potential (KN) was measured in the hydrogen and nitrogen atmospheres according to the decomposition of ammonia (NH3) in the nitriding atmosphere. The formation behavior of the compound layer was observed for various process conditions. In the nitriding atmosphere, the steel surface was oxidized to Fe3O4 and then reduced by NH3 to form a compound. During the initial nitriding, the γ′ (Fe4N) phase was formed inside owing to the concentration gradient after the formation of ε phase at a high KN. Both ε (Fe2–3N) and γ′ phases grow over time, and finally, a compound with three layers of γ′/ε/γ′ is formed. The formation mechanisms of these nitrided compounds were thermodynamically compared and analyzed using the CALPHAD (CALculation of PHAse Diagrams) method. Based on the modified Lehrer diagram, the formation mechanisms of the ε and γ′ phases during the nitriding process could be identified.

Axisymmetric base flow is investigated to understand flow physics associated with the massive flow separation at a subsonic speed. The detached-eddy simulation (DES) approach is well suited in the current separated flow with a known separation point. The upstream attached boundary layer is well represented with the Reynolds-averaged Navier–Stokes (RANS) mode, whereas the separated flow from the base is well captured in the large-eddy simulation (LES) mode. Since the spatial resolution in the LES zone impacts directly the fidelity of the DES computation, a systematic approach is applied to the computational grid. Current computational grids are designed for nearly isotropic grids in the separated region (i.e., LES zone) with much reduced anisotropy of the grid in the separating shear layer, compared to computational grids documented in literature. Current grids allow the separating shear layer to undergo the Kelvin–Helmholtz instability, resulting in a rapid shift from the RANS to LES mode right after the flow separation. In consequence, the axisymmetric base flow is well resolved in the current DES computation with good agreement to relevant experimental data including the mean base pressure and the center-line velocity in the wake. The base flow is further discussed with statistical data of the separated flow. Current DES simulation is also compared with a typical RANS simulation to emphasize the high fidelity of the computational approach.

As real-world prediction systems often face missing modalities at inference, incomplete multimodal learning (IML) remains a practical challenge. While prior methods aim to learn representations robust to missing inputs, representations from incomplete modalities inevitably deviate from their full-modality counterparts due to missing evidence. To explicitly leverage these deviations, we propose MARS (Missingness-Aware Residual-guided Specialization), a mixture-of-experts framework that guides expert specialization based on how representations are reshaped by missingness. By contrasting task representations derived from incomplete inputs with their complete counterparts during training, we derive a privileged residual signal that captures this representational gap. The residual signal guides a residual router to assign samples to experts specialized for the corresponding deviation patterns. In parallel, a feature router learns to imitate this routing behavior using only incomplete inputs, enabling deployment without access to full modalities. To mitigate this train-test router gap, we develop a discrepancy-aware noise regularization that adaptively perturbs the residual router's decisions when the feature router deviates, enhancing expert robustness under imperfect imitation. Experiments on multimodal classification (CASIA-SURF, CREMA-D, UPMC Food-101) and segmentation (MCubeS) under missing scenarios show that MARS consistently surpasses baselines while remaining efficient and extensible to diverse backbones and tasks.

Various neuroimaging studies suffer from small sample size problem which often limit their reliability. Meta-analysis addresses this challenge by aggregating findings from different studies to identify consistent patterns of brain activity. However, traditional approaches based on keyword retrieval or linear mappings often overlook the rich hierarchical structure in the brain. In this work, we propose a novel framework that leverages hyperbolic geometry to bridge the gap between neuroscience literature and brain activation maps. By embedding text from research articles and corresponding brain images into a shared hyperbolic space via the Lorentz model, our method captures both semantic similarity and hierarchical organization inherent in neuroimaging data. In the hyperbolic space, our method performs multi-level neuroimaging meta-analysis (MNM) by 1) aligning brain and text embeddings for semantic correspondence, 2) guiding hierarchy between text and brain activations, and 3) preserving the hierarchical relationships within brain activation patterns. Experimental results demonstrate that our model outperforms baselines, offering a robust and interpretable paradigm of multi-level neuroimaging meta-analysis via hyperbolic brain-text representation.

It seems as though progressively more people are in the race to upload content, data, and information online; and hospitals haven’t neglected this trend either. Hospitals are now at the forefront for multi-site medical data sharing to provide ground-breaking advancements in the way health records are shared and patients are diagnosed. Sharing of medical data is essential in modern medical research. Yet, as with all data sharing technology, the challenge is to balance improved treatment with protecting patient’s personal information. This paper provides a novel split learning algorithm coined the term, “multi-site split learning”, which enables a secure transfer of medical data between multiple hospitals without fear of exposing personal data contained in patient records. It also explores the effects of varying the number of end-systems and the ratio of data-imbalance on the deep learning performance. A guideline for the most optimal configuration of split learning that ensures privacy of patient data whilst achieving performance is empirically given. We argue the benefits of our multi-site split learning algorithm, especially regarding the privacy preserving factor, using CT scans of COVID-19 patients, X-ray bone scans, and cholesterol level medical data.