自适应融合层级特征的混合退化图像复原算法

多种退化类型混合的图像比单一类型的退化图像降质更严重,很难建立精确模型对其复原,研究端到端的神经网络算法是复原的关键.现有的基于操作选择注意力网络的算法(operation-wiseattentionnetwork,OWAN)虽然有一定的性能提升,但是其网络过于复杂,运行较慢,复原图像缺乏高频细节,整体效果也有提升的空间.针对这些问题,提出一种基于层级特征融合的自适应复原算法.该算法直接融合不同感受野分支的特征,增强复原图像的结构;用注意力机制对不同层级的特征进行动态融合,增加模型的自适应性,降低了模型冗余;另外,结合L1损失和感知损失,增强了复原图像的视觉感知效果.在DIV2K,BSD500等数据集上的实验结果表明,该算法无论是在峰值信噪比和结构相似性上的定量分析,还是在主观视觉质量方面,均优于OWAN算法,充分证明了该算法的有效性.     

Approaching Ultrastable High‐Rate Li–S Batteries through Hierarchically Porous Titanium Nitride Synthesized by Multiscale Phase Separation

Porous architectures are important in determining the performance of lithium–sulfur batteries (LSBs). Among them, multiscale porous architecutures are highly desired to tackle the limitations of single‐sized porous architectures, and to combine the advantages of different pore scales. Although a few carbonaceous materials with multiscale porosity are employed in LSBs, their nonpolar surface properties cause the severe dissolution of lithium polysulfides (LiPSs). In this context, multiscale porous structure design of noncarbonaceous materials is highly required, but has not been exploited in LSBs yet because of the absence of a facile method to control the multiscale porous inorganic materials. Here, a hierarchically porous titanium nitride (h‐TiN) is reported as a multifunctional sulfur host, integrating the advantages of multiscale porous architectures with intrinsic surface properties of TiN to achieve high‐rate and long‐life LSBs. The macropores accommodate the high amount of sulfur, facilitate the electrolyte penetration and transportation of Li⁺ ions, while the mesopores effectively prevent the LiPS dissolution. TiN strongly adsorbs LiPS, mitigates the shuttle effect, and promotes the redox kinetics. Therefore, h‐TiN/S shows a reversible capacity of 557 mA h g^?1 even after 1000 cycles at 5 C rate with only 0.016% of capacity decay per cycle.     

When Does Communication Learning Need Hierarchical Multi-Agent Deep Reinforcement Learning

Abstract

Multi-agent systems need to communicate to coordinate a shared task. We show that a recurrent neural network (RNN) can learn a communication protocol for coordination, even if the actions to coordinate are performed steps after the communication phase. We show that a separation of tasks with different temporal scale is necessary for successful learning. We contribute a hierarchical deep reinforcement learning model for multi-agent systems that separates the communication and coordination task from the action picking through a hierarchical policy. We further on show, that a separation of concerns in communication is beneficial but not necessary. As a testbed, we propose the Dungeon Lever Game and we extend the Differentiable Inter-Agent Learning (DIAL) framework. We present and compare results from different model variations on the Dungeon Lever Game.     

Hierarchical Coherent Phonons in a Superatomic Semiconductor

The coupling of phonons to electrons and other phonons plays a defining role in material properties, such as charge and energy transport, light emission, and superconductivity. In atomic solids, phonons are delocalized over the 3D lattice, in contrast to molecular solids where localized vibrations dominate. Here, a hierarchical semiconductor that expands the phonon space by combining localized 0D modes with delocalized 2D and 3D modes is described. This material consists of superatomic building blocks (Re₆Se₈) covalently linked into 2D sheets that are stacked into a layered van der Waals lattice. Using transient reflectance spectroscopy, three types of coherent phonons are identified: localized 0D breathing modes of isolated superatom, 2D synchronized twisting of superatoms in layers, and 3D acoustic interlayer deformation. These phonons are coupled to the electronic degrees of freedom to varying extents. The presence of local phonon modes in an extended crystal opens the door to controlling material properties from hierarchical phonon engineering.     

Synthesis of Hierarchical 4H/fcc Ru Nanotubes for Highly Efficient Hydrogen Evolution in Alkaline Media

Hierarchical metal nanostructures containing 1D nanobuilding blocks have stimulated great interest due to their abundant active sites for catalysis. Herein, hierarchical 4H/face‐centered cubic (fcc) Ru nanotubes (NTs) are synthesized by a hard template‐mediated method, in which 4H/fcc Au nanowires (NWs) serve as sacrificial templates which are then etched by copper ions (Cu²⁺) in dimethylformamide. The obtained hierarchical 4H/fcc Ru NTs contain ultrathin Ru shells (5–9 atomic layers) and tiny Ru nanorods with length of 4.2 ± 1.1 nm and diameter of 2.2 ± 0.5 nm vertically decorated on the surface of Ru shells. As an electrocatalyst for the hydrogen evolution reaction in alkaline media, the hierarchical 4H/fcc Ru NTs exhibit excellent electrocatalytic performance, which is better than 4H/fcc Au‐Ru NWs, commercial Pt/C, Ru/C, and most of the reported electrocatalysts.     

Progress in the design of zeolite catalysts for biomass conversion into biofuels and bio-based chemicals

This article reviews the recent advances in the development of zeolite catalysts for biomass valorization processes to produce both biofuels and/or bio-based chemicals, which is an emerging and fast expanding field. The work deals with different types of feedstocks, including vegetable oils, lignocellulose and sugars, as well as with a number of relevant intermediates and platform molecules. Transformation of biomass into valuable products is hindered by a number of factors, mainly related to its complex composition, as biomass typically consists of bulky molecules with high oxygen content. Accordingly, biomass processing usually requires the combination of multiple steps and severe conditions, hence concepts like atom efficiency, product selectivity, and catalyst deactivation become of special relevance. A great progress has been achieved in the past years engineering the properties of zeolites for being adapted to the challenges associated to biomass valorization. The possibility of tailoring the main physicochemical properties of zeolites has become now a reality, being the major reason that explains the success achieved by this class of materials in a growing variety of biomass conversion pathways, as those described in this work: catalytic cracking and pyrolysis, hydrotreatments, with special relevance for hydrodeoxygenation processes, as well as in a high number of condensation, isomerization, and dehydration reactions. Thus, the development of hierarchical zeolites, exhibiting enhanced accessibility, and the possibility of introducing and combining in a controlled way different types of active sites (Brønsted and Lewis acid centers, basic sites, and metal phases) are the main basis of the excellent performance of zeolites in numerous biomass conversion routes.     

重庆主城区城市空间结构演变与优化——基于公共服务功能组织视角

城市空间结构与城市公共服务功能组织密切相关。重庆主城区作为典型的"多中心、组团式"城市,在城市公共服务功能的就近组织、公共服务需求的均衡化方面具有先天的优势。本文梳理重庆主城区空间结构的演进历程以及与公共服务需求变化的互动关系,辨识随着城市功能由生产向消费转型,空间面临的结构问题及其内在动因,借鉴马斯洛理论,提出顺应当前及未来公共服务需求层次的分异,城市空间结构应从"多中心、组团式"进一步向"分层多中心、组团式"优化,引导形成相对集中、功能综合的基层"小、微中心"体系,以有序组织公共服务功能,最大限度地减少非必要交通出行,降低居民出行的时耗,提升城市效率,并借鉴恩格尔系数构建衡量城市效率的评价框架,为破解山地城市空间结构问题,推动城市空间转型提供新的思考视角。     

Three-dimensional porous SnO2@NC framework for excellent energy conversion and storage

SnO₂-based materials are deemed to be attractive electrodes for lithium/sodium ion batteries (LIBs and SIBs) and electrocatalytic CO₂ reduction reaction (CRR) because of high energy density and large abundance. However, the practical application of the SnO₂-based materials is prevented by low electrical conductivity and large volume change. Herein, we construct a three-dimensional (3D) porous network with SnO₂ nanoparticles into N-doped carbon (namely P–SnO₂@NC) synthesized by freeze drying followed by a pyrolyzation process. In the composite, the 3D hierarchical framework can facilitate the ion penetration and gas diffusion. In addition, the NC network can optimize the conductivity of the material and suppress the electrode material to fall off from the electrode. Therefore, the electrode delivers excellent electrochemical properties with high capacities of 510 mA h g⁻¹ after 1000 cycles for LIBs and 497 mA h g⁻¹ after 500 cycles for SIBs. Furthermore, the electrode shows high selectivity for CRR with a large coulombic efficiency (CE) of 52.7% for HCOOH at 0.6 V.     

机械结构分层优化技术研究进展

复杂机械结构优化设计一般表现为多目标、多约束、多参数的优化问题,所以复杂结构优化设计过程通常存在计算复杂、不易收敛等困难。分层优化技术是复杂结构优化设计的一种有效途径,通过将优化问题中复杂的约束、设计变量以及功能目标合理分解为若干子层问题进行求解,然后通过协调得到复杂机械结构整体优化问题的结果。文章对机械结构分层优化技术的理论、应用研究现状进行了总结并探讨了其关键技术环节及发展趋势。