Optimal Architecture Research Articles

U-Net and Its Variants Based Automatic Tracking of Radial Artery in Ultrasonic Short-Axis Views: A Pilot Study.

Background/Objectives: Radial artery tracking (RAT) in the short-axis view is a pivotal step for ultrasound-guided radial artery catheterization (RAC), which is widely employed in various clinical settings. To eliminate disparities and lay the foundations for automated procedures, a pilot study was conducted to explore the feasibility of U-Net and its variants in automatic RAT. Methods: Approved by the institutional ethics committee, patients as potential RAC candidates were enrolled, and the radial arteries were continuously scanned by B-mode ultrasonography. All acquired videos were processed into standardized images, and randomly divided into training, validation, and test sets in an 8:1:1 ratio. Deep learning models, including U-Net and its variants, such as Attention U-Net, UNet++, Res-UNet, TransUNet, and UNeXt, were utilized for automatic RAT. The performance of the deep learning architectures was assessed using loss functions, dice similarity coefficient (DSC), and Jaccard similarity coefficient (JSC). Performance differences were analyzed using the Kruskal-Wallis test. Results: The independent datasets comprised 7233 images extracted from 178 videos of 135 patients (53.3% women; mean age: 41.6 years). Consistent convergence of loss functions between the training and validation sets was achieved for all models except Attention U-Net. Res-UNet emerged as the optimal architecture in terms of DSC and JSC (93.14% and 87.93%), indicating a significant improvement compared to U-Net (91.79% vs. 86.19%, p < 0.05) and Attention U-Net (91.20% vs. 85.02%, p < 0.05). Conclusions: This pilot study validates the feasibility of U-Net and its variants in automatic RAT, highlighting the predominant performance of Res-UNet among the evaluated architectures.

Generative adversarial networks for whispered to voiced speech conversion: a comparative study

AbstractGenerative Adversarial Networks (GANs) have demonstrated promising results as end-to-end models for whispered to voiced speech conversion. Leveraging non-autoregressive systems like GANs capable of performing conditional waveform generation eliminates the need for separate models to estimate voiced speech features, and leads to faster inference compared to autoregressive methods. This study aims to identify the optimal GAN architecture for the whispered to voiced speech conversion task by comparing six state-of-the-art models. Furthermore, we present a method for evaluating the preservation of speaker identity and local accent, using embeddings obtained from speaker- and language identification systems. Our experimental results show that building the speech conversion system based on the HiFi-GAN architecture yields the best objective evaluation scores, outperforming the baseline by $$\sim$$ ∼ 9% relative using frequency-weighted Signal-to-Noise Ratio and Log Likelihood Ratio, as well as by $$\sim$$ ∼ 29% relative using Root Mean Squared Error. In subjective tests, HiFi-GAN yielded a mean opinion score of 2.9, significantly outperforming the baseline with a score of 1.4. Furthermore, HiFi-GAN enhanced ASR performance and preserved speaker identity and accent, with correct language detection rates of up to $$\sim$$ ∼ 98%.

Vision-Based Localization Method for Picking Points in Tea-Harvesting Robots.

To address the issue of accurately recognizing and locating picking points for tea-picking robots in unstructured environments, a visual positioning method based on RGB-D information fusion is proposed. First, an improved T-YOLOv8n model is proposed, which improves detection and segmentation performance across multi-scale scenes through network architecture and loss function optimizations. In the far-view test set, the detection accuracy of tea buds reached 80.8%; for the near-view test set, the mAP0.5 values for tea stem detection in bounding boxes and masks reached 93.6% and 93.7%, respectively, showing improvements of 9.1% and 14.1% over the baseline model. Secondly, a layered visual servoing strategy for near and far views was designed, integrating the RealSense depth sensor with robotic arm cooperation. This strategy identifies the region of interest (ROI) of the tea bud in the far view and fuses the stem mask information with depth data to calculate the three-dimensional coordinates of the picking point. The experiments show that this method achieved a picking point localization success rate of 86.4%, with a mean depth measurement error of 1.43 mm. The proposed method improves the accuracy of picking point recognition and reduces depth information fluctuations, providing technical support for the intelligent and rapid picking of premium tea.

Semantic Segmentation Uncertainty Assessment of Different U-net Architectures for Extracting Building Footprints

Abstract. Automatic extraction of building footprints from aerial and space imageries has been found ever increasing importance in urban planning, disaster management, and environmental monitoring. However, achieving accurate building footprint extraction poses significant challenges due to diverse building characteristics and their similarities to their background elements. While conventional methods in building footprint extraction have mainly relied on image processing techniques, recent advancements in deep learning, particularly semantic segmentation algorithms like U-Net, have shown promise in addressing these challenges through machine learning. This study explores different depths of the U-Net model for building footprint extraction, aiming to identify the optimum architecture while investigating the semantic uncertainty of the building footprint extraction. Utilizing aerial imagery from cities including Berlin, Paris, Chicago, and Zurich, collected from Google Maps and OpenStreetMap (OSM) data, five U-Net models have been compared with varying depths. In addition, the impact of dataset sizes and learning rates on model performance has been investigated. Results confirmed that the U-Net-32-1024 model achieves the highest intersection over union (IoU), Accuracy, and F1-score. Moreover, increasing the training dataset size leads to significant improvements in model performance with IoU, Accuracy and F1-score reaching their values of 73.73%, 88.65% and 88.53%. However, challenges remain in accurately delineating buildings in dense urban areas. Nonetheless, our findings demonstrated the effectiveness of U-Net models in building footprint extraction.

Design and Application of Intelligent Recommender Systems in Vocational Rehabilitation

With the rapid development of information technology, the application of intelligent recommender systems in various fields has become more and more extensive, and vocational rehabilitation, as a field focusing on personalized services, faces the challenge of how to effectively use big data and intelligent algorithms to enhance the rehabilitation effect. This paper discusses the design principles and key technologies of intelligent recommender system in vocational rehabilitation, combining data-driven recommendation algorithms with users' personalized needs to achieve accurate recommendation of vocational rehabilitation programs. Through the optimization of system architecture and functional implementation, the intelligent recommender system shows remarkable effects in practical application, effectively improving the decision-making efficiency and user satisfaction in the rehabilitation process. Finally, this paper gives an outlook on the future development of the intelligent recommendation system in the field of vocational rehabilitation, and proposes possible research directions and application expansion paths.

Quantum-Resilient Self-Evolving Blockchains: AI- Driven Consensus and Autonomous Security Upgrades

The fragility of conventional blockchain systems is becoming more apparent in face of the relentless march forward in quantum computing. In this paper, we suggest the optimal architecture for quantum-resistant blockchains deploying AI regarding consensus self- evolving security protocols and mechanisms. An intuitive example of this integration is the use of machine learning algorithms for a self-adapting security system, which automatically adjusts the protection on its infrastructure capabilities to evolving quantum threats. In this paper, we address the consequences of quantum computing in blockchain security introduce a framework for AI-assisted consensus (Section 3) and cover the feasibility of self- evolving blockchains that guarantee their reliability as well as integrity in a post-quantum era.

Effect of crossing warp arrangements on delamination resistance of 3D woven composite T-joints under in-plane tensile loading

An experimental study for investigating the delamination behaviour of 3D woven composite T-joints with weave variations and optimising weave architectures is carried out. This study involves 10 types of crossing warp architectures at the junction. Quasi-static tensile load is applied to two flanges of 3D woven composite T-joints to evaluate the in-plane mechanical performance. The crossing warp architecture effectively improves the in-plane mechanical performance. Results indicate a significant influence of crossing warp arrangements on failure modes of the 3D woven composite T-joints. The use of internal crossing warp architectures leads to severe delamination in the 3D woven composite T-joints while the composite T-joints with 3D woven external crossing warps primarily fail due to the debonding of fibres and matrix and fibre breakage. The optimal weave architecture for 3D woven composite T-joints is confirmed by analysing the in-plane mechanical behaviour with different crossing warp arrangements and proportions. Regardless of the crossing warp proportions, the external crossing warp architectures outperformed their internal counterparts in resisting delamination, resulting in a maximum increase of 68.75 %, 30.04 % and 116.81 % in modulus, strength and failure strain respectively.

A Trianalyte µPAD for Simultaneous Determination of Iron, Zinc, and Manganese Ions.

In this work, a microfluidic paper-based analytical device (µPAD) for simultaneous detection of Fe, Zn, and Mn ions using immobilized chromogenic reagents Ferene S, xylenol orange, and 1-(2-pyridylazo)-2-naphthol, respectively, is presented. As the effective recognition of analytes via respective chromogens takes place under extremely different pH conditions, experiments reported in this publication are focused on optimization of the µPAD architecture allowing for the elimination of potential cross effects. The paper-based microfluidic device was fabricated using low-cost and well-reproducible wax-printing technology. For optical detection of color changes, an ordinary office scanner and self-made RGB-data processing program were applied. Optimized and stable over time, µPADs allow fast, selective, and reproducible multianalyte determinations at submillimolar levels of respective heavy metal ions, which was confirmed by results of the analysis of solutions mimicking real samples of wastewater. The presented concept of simultaneous determination of different analytes that required extremely different conditions for detection can be useful for the development of other multianalyte microfluidic paper-based devices in the µPAD format.

Long-distance photon-mediated and short-distance entangling gates in three-qubit quantum dot spin systems

Superconducting resonator couplers will likely become an essential component in modular semiconductor quantum dot (QD) spin qubit processors, as they help alleviate crosstalk and wiring issues as the number of qubits increases. Here, we focus on a three-qubit system composed of two modules: a two-electron triple QD resonator coupled to a single-electron double QD. Using a combination of analytical techniques and numerical results, we derive an effective Hamiltonian that describes the three-qubit logical subspace and show that it accurately captures the dynamics of the system. We examine the performance of short-range and long-range entangling gates, revealing the effect of a spectator qubit in reducing the gate fidelities in both cases. We further study the competition between nonadiabatic errors and spectator-associated errors in short-range operations and quantify their relative importance across practical parameter ranges for short and long gate times. We also analyze the impact of charge noise together with residual coupling to the spectator qubit on intermodule entangling gates and find that for current experimental settings, leakage errors are the main source of infidelities in these operations. Our results help pave the way toward identifying optimal modular QD architectures for quantum information processing on semiconductor chips. Published by the American Physical Society 2024

Selective and multi-scale fusion Mamba for medical image segmentation

Given the high variability in the morphology and size of lesion areas in medical images, accurate medical image segmentation requires both precise positioning of global contours and careful processing of local boundaries. This emphasizes the importance of fusing multi-scale local and global features. However, existing CNN and Transformer-based models are often limited by high parameter counts and complex calculations, making it difficult to efficiently integrate these features. To address this challenge, we proposed two innovative optimization architectures: Selective Fusion Mamba (SF-Mamba) and Multi-Scale Fusion Mamba (MF-Mamba). SF-Mamba can flexibly and dynamically adjust the fusion strategy of local and global features according to the characteristics of the lesions, effectively handling segmentation tasks with variable morphology. MF-Mamba enhances the model’s segmentation ability for lesions of different sizes by capturing global information across scales. Based on these two structures, we constructed a lightweight SMM-UNet model, which not only significantly reduces the computational burden (with only 0.038M parameters) but also demonstrates excellent generalization ability and can efficiently adapt to various types of medical images. Extensive tests on the ISIC2017, ISIC2018, and BUSI public datasets show that SMM-UNet achieves excellent segmentation performance with an extremely low parameter cost.

Determination of the Optimum Architecture of Additively Manufactured Magnetic Bioactive Glass Scaffolds for Bone Tissue Engineering and Drug-Delivery Applications.

For better bone regeneration, precise control over the architecture of the scaffolds is necessary. Because the shape of the pore may affect the bone regeneration, therefore, additive manufacturing has been used in this study to fabricate magnetic bioactive glass (MBG) scaffolds with three different architectures, namely, grid, gyroid, and Schwarz D surface with 15 × 15 × 15 mm3 dimensions and 70% porosity. These scaffolds have been fabricated using an in-house-developed material-extrusion-based additive manufacturing system. The composition of bioactive glass was selected as 45% SiO2, 20% Na2O, 23% CaO, 6% P2O5, 2.5% B2O3, 1% ZnO, 2% MgO, and 0.5% CaF2 (wt %), and additionally 0.4 wt % of iron carbide nanoparticles were incorporated. Afterward, MBG powder was mixed with a 25% (w/v) Pluronic F-127 solution to prepare a slurry for fabricating scaffolds at 23% relative humidity. The morphological characterization using microcomputed tomography revealed the appropriate pore size distribution and interconnectivity of the scaffolds. The compressive strengths of the fabricated grid, gyroid, and Schwarz D scaffolds were found to be 14.01 ± 1.01, 10.78 ± 1.5, and 12.57 ± 1.2 MPa, respectively. The in vitro study was done by immersing the MBG scaffolds in simulated body fluid for 1, 3, 7, and 14 days. Darcy's law, which describes the flow through porous media, was used to evaluate the permeability of the scaffolds. Furthermore, an anticancer drug (Mitomycin C) was loaded onto these scaffolds, wherein these scaffolds depicted good release behavior. Overall, gyroid-structured scaffolds were found to be the most suitable among the three scaffolds considered in this study for bone tissue engineering and drug-delivery applications.

Machine learning-based analytical approach for mechanical analysis of composite hydrogen storage tanks under internal pressure

This study is a practical exploration of the application of machine learning for the mechanical analysis of filament-wound thin-composite hydrogen storage tanks under internal pressure. Our innovative approach seamlessly integrates classical laminate theory, comprehensive parametric analysis, and machine learning to advance the state-of-the-art in composite tank design. This versatile framework holds promise for addressing challenges in other structurally complex systems. A comprehensive parametric study was then conducted to explore the influence of key design parameters, such as internal pressure, tank radius, order of layer, and layer orientation. Pearson's correlation analysis was used to determine the parameters that had the most significant impact on the mechanical behavior of the tank. In addition, five machine learning models, namely, Extreme Gradient Boosting, Adaptive Boosting, Artificial Neural Network, Support Vector Machine, and Random Forest, were examined to determine their predictive capability for the mechanical performance of these composite tanks. Key findings demonstrate XGBoost's superior predictive capabilities, achieving a remarkable 99% accuracy in predicting stress in the x-direction compared to Random Forest's 96%. XGBoost also outperformed Random Forest in terms of Mean Absolute Error (MAE) and Mean Absolute Percentage Error (MAPE) with values of 26 and 0.02, respectively, versus 207 and 0.16. These results underscore XGBoost's potential to revolutionize hydrogen storage tank design. The development of a user-friendly GUI application further enhances the practicality of this research, allowing engineers and composite designers to efficiently simulate the mechanical behavior of these tanks in real time by adjusting the cursors for variables such as the pressure, radius, layer order, and layer orientation. This tool, with its intuitive interface and elimination of manual calculations, promises significant enhancements in the design process, providing a platform for the rapid assessment and optimization of tank architectures, thereby improving the efficiency and reliability of the design process.

Real-time robust liver and gallbladder segmentation during laparoscopic cholecystectomy using convolutional neural networks: an analysis

Aim: Images in different laparoscopic cholecystectomy datasets are acquired using various camera models, parameters, and settings, with the annotation methods varying by institution. These factors result in inconsistent inference performance of the network model. This study aims to identify the optimal network model architecture for liver and gallbladder segmentation from several options. Then, the performance and robustness of the optimal network model are evaluated using an independent dataset that is not included in the training. Methods: The public dataset, CholecSeg8k, was utilized as the input for the network model training, validation, and testing. A local private dataset from KPJ Damansara Hospital, Selangor, Malaysia, was used for testing purposes only. For the implementation of liver and gallbladder segmentation, segmentation models, a public Python library was employed. Results: Among the experiments, highly accurate liver and gallbladder segmentation results were achieved using the feature pyramid network (FPN) architecture as the network model, with the Inception-ResNet-v2 architecture as the network backbone. The best-trained network model resulted in a loss of 0.070955, a mean intersection over union (IoU) score of 0.95896, and a mean F1-score of 0.9773 on the test set. However, visualized results for the private dataset contained considerable false-negative areas. Conclusion: The proposed automated technique has the potential to serve as an alternative to the conventional indocyanine green injection along with near-infrared fluorescence imaging (ICG-NIRF)-based method for liver and gallbladder segmentation during laparoscopic cholecystectomy. Future work will focus on enhancing the results of the private dataset. Additionally, a surgeon-assistant robotic arm that will use the liver and gallbladder segmentation results for camera steering will be analyzed.

Study on Optimization of Urban Architectural Design Based on Public Demands

As the material standard of living continues to improve, people's needs at a higher level have triggered a new craze. The change of view on architecture has become an intuitive perception of this boom in people's lives, and the call for architectural design to respond to public demands is increasing. This paper compares the design and practical effects of the Amsterdam Library and the Tencent Building in Chengdu for public demands, combines the new demands of the public for modern architecture to conduct a literature analysis of relevant cases, and then explores the direction and strategy of modern architectural design and optimization oriented by public demands. It is found that the physiological, psychological, and health values provided by buildings are desired by the public but not fully satisfied, and that multifunctionality, aesthetic experience, green intelligence, and humanization are the mainstream of future buildings desired by people. Refining and grasping the information of public demands and using it as a reference is an important part of reaching the transformation of architectural design. In urban building construction, the use of the cycle is generally longer, the current reality is to minimize the misalignment of supply and demand, which means that must do a good job of mastering the information of public demands and have a certain prediction of its future direction, reserved for the transformation of the elastic space, with the times.

Automated Recognition of Snow-Covered and Icy Road Surfaces Based on T-Net of Mount Tianshan

The Tianshan Expressway plays a crucial role in China’s “Belt and Road” strategy, yet the extreme climate of the Tianshan Mountains poses significant traffic safety risks, hindering local economic development. Efficient detection of hazardous road surface conditions (RSCs) is vital to address these challenges. The complexity and variability of RSCs in the region, exacerbated by harsh weather, make traditional surveillance methods inadequate for real-time monitoring. To overcome these limitations, a vision-based artificial intelligence approach is urgently needed to ensure effective, real-time detection of dangerous RSCs in the Tianshan road network. This paper analyzes the primary structures and architectures of mainstream neural networks and explores their performance for RSC recognition through a comprehensive set of experiments, filling a research gap. Additionally, T-Net, specifically designed for the Tianshan Expressway engineering project, is built upon the optimal architecture identified in this study. Leveraging the split-transform-merge structure paradigm and asymmetric convolution, the model excels in capturing detailed information by learning features across multiple dimensions and perspectives. Furthermore, the integration of channel, spatial, and multi-head attention modules enhances the weighting of key features, making the T-Net particularly effective in recognizing the characteristics of snow-covered and icy road surfaces. All models presented in this paper were trained on a custom RSC dataset, compiled from various sources. Experimental results indicate that the T-Net outperforms fourteen once state-of-the-art (SOTA) models and three models specifically designed for RSC recognition, with 97.44% accuracy and 9.79% loss on the validation set.

Implementation of Artificial Neural Networks on Field-Programmable Gate Arrays

Implementing Artificial Neural Networks (ANNs) on Field-Programmable Gate Arrays (FPGAs) provides a promising solution for achieving high-performance, low-latency, and energy-efficient computations in complex tasks. This paper investigates the methodology for mapping ANNs onto FPGAs, focusing on critical aspects such as architecture selection, hardware design, and optimization techniques. By harnessing the parallel processing capabilities and reconfigurability of FPGAs, neural network computations are significantly accelerated, making them ideal for real-time applications like image processing and embedded systems. The implementation process addresses key considerations, including fixed-point arithmetic, memory management, and dataflow optimization, while employing advanced techniques such as pipelining, quantization, and pruning. The research compares the accuracy and performance speedup of ANNs on CPUs versus FPGAs, revealing that FPGA-based simulations are 4680 times faster than CPU-based simulations using MATLAB, without compromising prediction accuracy.

AutoNet-Generated Deep Layer-Wise Convex Networks for ECG Classification.

The design of neural networks typically involves trial-and-error, a time-consuming process for obtaining an optimal architecture, even for experienced researchers. Additionally, it is widely accepted that loss functions of deep neural networks are generally non-convex with respect to the parameters to be optimised. We propose the Layer-wise Convex Theorem to ensure that the loss is convex with respect to the parameters of a given layer, achieved by constraining each layer to be an overdetermined system of non-linear equations. Based on this theorem, we developed an end-to-end algorithm (the AutoNet) to automatically generate layer-wise convex networks (LCNs) for any given training set. We then demonstrate the performance of the AutoNet-generated LCNs (AutoNet-LCNs) compared to state-of-the-art models on three electrocardiogram (ECG) classification benchmark datasets, with further validation on two non-ECG benchmark datasets for more general tasks. The AutoNet-LCN was able to find networks customised for each dataset without manual fine-tuning under 2 GPU-hours, and the resulting networks outperformed the state-of-the-art models with fewer than 5% parameters on all the above five benchmark datasets. The efficiency and robustness of the AutoNet-LCN markedly reduce model discovery costs and enable efficient training of deep learning models in resource-constrained settings.

A Generalized Learning Approach to Deep Neural Networks

Optimization of machine learning architectures is essential in determining the efficacy and the applicability of any neural architecture to real world problems. In this work a generalized Newton's method (GNM) is presented as a powerful approach to learning in deep neural networks (DNN). This technique was compared to two popular approaches, namely the stochastic gradient descent (SGD) and the Adam algorithm, in two popular classification tasks. The performance of the proposed approach confirmed it as an attractive alternative to state-of-the-art first order solutions. Due to the good results presented in the case of shallow DNN, in the last part of the article an hybrid optimization method is presented. This method consists in combining two optimization algorithms, i.e. GNM and Adam or GNM and SGD, during the training phase within the layers of the neural network. This configuration aims to benefit from the strengths of both first- and second-order algorithms. In this case a convolutional neural network is considered and its parameters are updated with a different optimization algorithm. Also in this case, the hybrid approach returns the best performance with respect to the first order algorithms.

Mechanics of Materials: Multiscale Design of Advanced Materials and Structures

Materials can now be designed and architectured like structural components for targeted mechanical and physical properties. Structures and microstructures should not be studied independently and their design will benefit from a multiscale approach combining nonlinear continuum mechanics approaches and physical descriptions of elasticity, viscoplasticity, phase transformations and damage of microstructures, at various scales. The aim of the workshop was to gather outstanding junior and senior researchers in the various branches of mathematics, physics and engineering sciences suited to address the question of design of materials and structures by means of multiscale discrete and continuum approaches to their constitutive behavior. Examples include atomic or macroscopic lattices, random or periodic cellular materials, smart materials like shape memory alloys, 3D woven composites, acoustic and electromagnetic metamaterials, etc. Modern continuum mechanics relies on sophisticated constitutive laws for anisotropic materials exhibiting elastoviscoplastic behavior, still a field of intense research with new mathematical concepts. In particular size-dependent properties are addressed by resorting to generalized continua such as gradient or micromorphic and phase field models. The latter are attractive for the simulation of microstructure evolution coupled with mechanics, due to thermodynamic and metallurgical processes and damage. Scale transition and homogenization methods for continuous and discrete systems are required for the determination of effective material and structural behavior. Metamaterials are architectured materials specifically designed to achieve certain propagation and dispersion properties of elastic and plastic waves. Optimization strategies for the design of optimal architectures are involved in the design process. Target functions for optimization are now based on multicriteria (stiffness, strength, thermal expansion, transport properties, anisotropy etc.).

Electroluminescent and photoluminescent light-emitting diodes from carbon dots and device architecture optimization.

Light emission from carbon dots (CDs) is of great interest in both electroluminescence and photoluminescence. Herein, we construct electroluminescent and photoluminescent light-emitting diodes (LEDs) from emitters of CDs. The electroluminescent LED with host-guest-doped dual emissive layers (EMLs) of [poly(9-vinylcarbazole) (PVK)-CDs] × 2 gives satisfactory electro-optical properties, with maximum luminance of 560 cd m-2 at 16 V, luminous efficiency of 0.183 cd A-1, power efficiency of 0.082 lm W-1, and external quantum efficiency of 0.25%, which are superior to the counterparts with single-EML of CDs, single-EML of [PVK-CDs], and triple-EMLs of [PVK-CDs] × 3. These enhanced properties are rationally ascribed to optimization of the device architecture, carrier balance improvement, and reduction in concentration-induced quenching. The electroluminescent LEDs also show color evolution from PVK to CDs, and/or to the PVK/CDs interface, with increasing driving voltages, owing to incomplete energy transfer from PVK to CDs. Highly efficient photoluminescent LEDs with 365- and 395-nm UV excitation are demonstrated. With a CDs : polyvinyl pyrrolidone ratio of 1 : 3, the 365-nm excited photoluminescent LED gives a maximum luminance of 512 347 cd m-2 with a power efficiency of 25.2 lm W-1, while the 395-nm excited photoluminescent device gives a maximum luminance of 670 954 cd m-2 with a power efficiency of 22.0 lm W-1, with both showing yellow emission. Our experiments provide some new ideas for broadening CD applications and advancing LEDs.