Search

Here we present a new Explainable AI (XAI) method to probe the functional partition in AI models by comparing features attended to at different layers with human attention driven by diverse task demands. We applied this method to explain an object detector Yolo-v5s in multi-category and single-category object detection tasks. We found that the model's neck showed higher similarity to human attention during object detection, indicating a reliance on diagnostic features in the neck, whereas its backbone showed higher similarity to attention during passive viewing, indicating salient local features encoded. With this understanding of its functional partition, using Yolo-v5s as a model for human cognition, our comparative analysis against human attention when providing explanations for object detection revealed that humans attended to a combination of diagnostic and salient features during explaining multi-category general object detection but attended to mainly diagnostic features when explaining single-category human/vehicle detection in driving scenarios.

The effects of interface oxidation on the transport behavior of the 2-D electron gas (2DEG) in AlGaN/GaN heterostructures by plasma-enhanced-atomic-layer-deposited AlN (PEALD-AlN) passivation were investigated using temperature-dependent Hall-effect and X-ray photoelectron spectroscopy (XPS) characterizations. AlGaN/GaN heterostructure with a 4-nm-thick PEALD-AlN passivation exhibits good 2DEG transport behavior and stability at moderately high temperature (e.g., 275 °C). However, serious oxidation of the AlN/GaN (cap layer) interface occurs as the sample is heated up to 400 °C in low-pressure atmosphere, as verified by an increased Ga-O bond in Ga 3d core-level spectra. The oxidation leads to a significant reduction of 2.47 × 1012 cm-2 in the 2DEG density in the channel. A modified AlN passivation structure with Al2O3/AlN (10/4 nm) stack is shown to be able to effectively suppress the oxidation of the AlN/GaN interface, demonstrating an enhanced 2DEG density and high-temperature stability even when the sample is heated up to 500 °C. Based on XPS and 2DEG recovery experiments, it is suggested that acceptor-like deep levels have been generated in the near-surface region of AlGaN/GaN heterostructure because of the oxidation, and trapping of these deep levels results in significant depletion of the 2DEG in the channel. The effects of PEALD-AlN passivation on the strain in the AlGaN barrier of AlGaN/GaN heterostructures are also evaluated with high-resolution X-ray diffraction technique. [ABSTRACT FROM AUTHOR]

In this letter, we report high-performance enhancement-mode (E-mode) Al2O3/AlGaN/GaN metal–insulator–semiconductor high-electron-mobility transistors (MIS-HEMTs) fabricated with high-temperature low-damage gate recess technique. The high-temperature gate recess is implemented by increasing the substrate temperature to 180 °C to enhance the desorption of chlorine-based etching residues during the dry etching of AlGaN barrier. High-crystal-quality Al2O3 gate dielectric was grown by atomic-layer deposition using O3 as the oxygen source to suppress hydrogen-induced weak bonds. The fabricated E-mode MIS-HEMTs exhibit a threshold voltage of 1.6 V, a pulsed drive current of 1.13 A/mm, and very low OFF-state standby power of 6.8 \times 10^\mathrm -8 W/mm at V\mathrm {GS} =0 V and V\mathrm {DS} =30 V. At 4 GHz and in pulse-mode operation, the output power density and power-added efficiency were measured to be 5.76 W/mm and 57%, both of which are the highest for GaN-based E-mode MIS-HEMTs reported to date. [ABSTRACT FROM AUTHOR]

This study explores the integration of crop phenology models and machine learning (ML) approaches for predicting rice phenology across China, to gain a deeper understanding of rice phenology prediction. Multiple approaches were used to predict heading and maturity dates at 337 locations across the main rice growing regions of China from 1981 to 2020, including crop phenology model, machine learning and hybrid model that integrate both approaches. Furthermore, an interpretable machine learning (IML) using SHapley Additive exPlanation (SHAP) was employed to elucidate influence of climatic and varietal factors on uncertainty in crop phenology model predictions. Overall, the hybrid model demonstrated a high accuracy in predicting rice phenology, followed by machine learning and crop phenology models. The best hybrid model, based on a serial structure and the eXtreme Gradient Boosting (XGBoost) algorithm, achieved a root mean square error (RMSE) of 4.65 and 5.72 days and coefficient of determination (R 2 ) values of 0.93 and 0.9 for heading and maturity predictions, respectively. SHAP analysis revealed temperature to be the most influential climate variable affecting phenology predictions, particularly under extreme temperature conditions, while rainfall and solar radiation were found to be less influential. The analysis also highlighted the variable importance of climate across different phenological stages, rice cultivation patterns, and geographic regions, underscoring the notable regionality. The study proposed that a hybrid model using an IML approach would not only improve the accuracy of prediction but also offer a robust framework for leveraging data-driven in crop modeling, providing a valuable tool for refining and advancing the modeling process in rice., (Copyright © 2024. Published by Elsevier B.V.)