Learn decision trees with deep visual primitives

In this paper, we strive to propose a self-interpretable framework, termed PrimitiveTree, that incorporates deep visual primitives condensed from deep features with a conventional decision tree, bridging the gap between deep features extracted from deep neural networks (DNNs) and trees’ transparent decision-making processes. Specifically, we utilize a codebook, which embeds the continuous deep features into a finite discrete space (deep visual primitives) to distill the most common semantic information. The decision tree adopts the spatial location information and the mapped primitives to present the decision-making process of the deep features in a tree hierarchy. Moreover, the trained interpretable PrimitiveTree can inversely explain the constituents of the deep features, highlighting the most critical and semantic-rich image patches attributing to the final predictions of the given DNN. Extensive experiments and visualization results validate the effectiveness and interpretability of our method.     

计算机网络办公自动化及安全策略探究

互联网通讯采取标准化模式主要以TCP/IP协议为载体,通讯的优越特性体现在同时具备便捷性与开放性,为办公提供很大的便利,但基于网络系统也会入侵病毒、也会给信息数据与办公体系安全性造成威胁,直接影响企业综合稳定发展。据此,为保障办公工作的顺利开展,本文对计算机网络办公自动化及安全策略进行详细分析。     

Enhanced wide-activated residual network for efficient and accurate image deblocking

To save bandwidth and storage space as well as speed up data transmission, people usually perform lossy compression on images. Although the JPEG standard is a simple and effective compression method, it usually introduces various visually unpleasing artifacts, especially the notorious blocking artifacts. In recent years, deep convolutional neural networks (CNNs) have seen remarkable development in compression artifacts reduction. Despite the excellent performance, most deep CNNs suffer from heavy computation due to very deep and wide architectures. In this paper, we propose an enhanced wide-activated residual network (EWARN) for efficient and accurate image deblocking. Specifically, we propose an enhanced wide-activated residual block (EWARB) as basic construction module. Our EWARB gives rise to larger activation width, better use of interdependencies among channels, and more informative and discriminative non-linearity activation features without more parameters than residual block (RB) and wide-activated residual block (WARB). Furthermore, we introduce an overlapping patches extraction and combination (OPEC) strategy into our network in a full convolution way, leading to large receptive field, enforced compatibility among adjacent blocks, and efficient deblocking. Extensive experiments demonstrate that our EWARN outperforms several state-of-the-art methods quantitatively and qualitatively with relatively small model size and less running time, achieving a good trade-off between performance and complexity.     

建筑工业化发展的多群体演化博弈与仿真研究

新型建筑工业化具有高质量、低消耗、可循环发展等特征，其推广已上升到国家战略层面。利用演化博弈方法，建立“政府-开发商-银行” 的三方动态演化博弈模型，进行各博弈主体策略的演化稳定性分析，并针对初始状态、奖惩力度、借贷风险和开发成本等对演化结果的影响进行动态仿真。在此基础上， 考虑开发商群体的网络拓扑特征对演化真实性的影响，引入复杂网络理论， 以无标度网络为载体描述开发商个体的连接偏好和决策机制，构建政府监管下的建筑工业化扩散模型，并通过仿真深入研究相关因素对扩散深度的影响作用，最后结合仿真结果给出相应对策建议。     

Constructing magnetically recoverable CdS/CaFe2O4 P–N heterojunctions for enhanced photoelectrochemical properties towards H2 evolution reaction

A novel CdS/CaFe₂O₄ (CS/CFO) heterogeneous p-n junction was created by thermal deposition of CaFe₂O₄ nanoparticles on CdS rods. The CS/CFO hetero-structured photocatalysts exhibited increasingly efficient visible light harvesting compared to the bare CdS. The CS/CFO composites also presented higher photocurrent and slower decay of photoluminescence, suggesting a better separation of the photo-generated electrons and holes. The photocatalytic H₂ evolution quantity on the optimized CS/CFO composite from water in the presence of ethanol was up to 2200 μmol after 3-h visible light illumination, which is more than twice that of the pristine CdS. The chemical interaction between CdS and CaFe₂O₄ was confirmed by the shifts in the XPS peaks, which made it possible for the charge carriers to transfer across the p-n junction interface. This research highlights the importance of forming an interfacial p-n heterojunction between two semiconductors for efficient charge separation and improved photocatalytic performance.     

Graphitic-C3N4/ZnCr-layered double hydroxide 2D/2D nanosheet heterojunction: Mesoporous photocatalyst for advanced oxidation of azo dyes with in situ produced H2O2

In this study the constructional modification of Graphitic carbon nitride nanosheet (GCN-ns) has been made with the aid of ZnCr layered double hydroxide (ZC-LDH) in a unique 2D-2D structure to enhance its visible light absorption. Optical and morphological study presents successful incorporation of ZC-LDH on the surface of GCN-ns. Through adjusting of GCN-ns by ZC-LDH lower recombination rate of e^?/h⁺ pairs, longer lifetimes and an increase in contamination reduction was brought out. The binary nanocomposite was employed to effectively degrade Rhodamine B under UV/vis light irradiation. The improvement in photocatalytic abilities was proven to be related to in situ self-production of H₂O₂ on GCN-ns/ZC-LDH surface by Xe light irradiation which in return accounts for additional hydroxide radical generation. Radical quenching experiments specified the main active species involved while the consequent step-scheme (S-scheme) charge transfer mechanism was proposed.     

基于阶段注意力机制的电力线提取算法

电力系统维护是电力系统稳定运行的重要保障，应用智能算法的无人机电力巡检则为电力系统维护提供便捷。电力线提取是自主电力巡检以及保障飞行器低空飞行安全的关键技术，结合深度学习理论进行电力线提取是电力巡检的重要突破点。本文将深度学习方法用于电力线提取任务，结合电力线图像特点嵌入改进的图像输入策略和注意力模块，提出一种基于阶段注意力机制的电力线提取模型（SA-Unet）。本文提出的SA-Unet模型编码阶段采用阶段输入融合策略（Stage input fusion strategy， SIFS），充分利用图像的多尺度信息减少空间位置信息丢失。解码阶段通过嵌入阶段注意力模块（Stage attention module，SAM）聚焦电力线特征，从大量信息中快速筛选出高价值信息。实验结果表明，该方法在复杂背景的多场景中具有良好的性能。     

Total Site Hydrogen Integration with fresh hydrogen of multiple quality and waste hydrogen recovery in refineries

This paper proposes a novel method combining Pinch Methodology and waste hydrogen recovery, aiming to minimise fresh hydrogen consumption and waste hydrogen discharge. The method of multiple-level resource Pinch Analysis is extended to the level of Total Site Hydrogen Integration by considering fresh hydrogen sources with various quality. Waste hydrogen after Total Site Integration is further regenerated. The technical feasibility and economy of the various purification approaches are considered, demonstrated with a case study of a refinery hydrogen network in a petrochemical industrial park. The results showed that fresh hydrogen usage and waste hydrogen discharge could be reduced by 21.3% and 67.6%. The hydrogen recovery ratio is 95.2%. It has significant economic benefits and a short payback period for Total Site Hydrogen Integration with waste hydrogen purification. The proposed method facilitates the reuse of waste hydrogen before the purification process that incurs an additional environmental footprint. In line with the Circular Economy principles, hydrogen resource is retained in the system as long as possible before discharge.     

Synthesis of a Fe3O4-rGO-ZnO-catalyzed photo-Fenton system with enhanced photocatalytic performance

In this work, we designed a magnetically-separable Fe₃O₄-rGO-ZnO ternary catalyst, ZnO anchored on the surface of reduced graphene oxide (rGO)-wrapped Fe₃O₄ magnetic nanoparticles, where rGO, as an effective interlayer, can enhance the synergistic effect between ZnO and Fe₃O₄. The effects of three operational parameters, namely irradiation time, hydrogen peroxide dosage, and the catalyst dosage, on the photo-Fenton degradation of methylene blue and methyl orange were investigated. The results showed that the Fe₃O₄-rGO-ZnO had great potential for the destruction of organic compounds from wastewater using the Fenton chemical oxidation method at neutral pH. Repeatability of the photocatalytic activity after 5 cycles showed only a tiny drop in the catalytic efficiency.     

Generating Questions Based on Semi-Automated and End-to-End Neural Network

With the emergence of large-scale knowledge base, how to use triple information to generate natural questions is a key technology in question answering systems. The traditional way of generating questions require a lot of manual intervention and produce lots of noise. To solve these problems, we propose a joint model based on semi-automated model and End-to-End neural network to automatically generate questions. The semi-automated model can generate question templates and real questions combining the knowledge base and center graph. The End-to-End neural network directly sends the knowledge base and real questions to BiLSTM network. Meanwhile, the attention mechanism is utilized in the decoding layer, which makes the triples and generated questions more relevant. Finally, the experimental results on SimpleQuestions demonstrate the effectiveness of the proposed approach.