Current Chinese Science - Volume 3, Issue 2, 2023

Introduction: Representations play an essential role in learning artificial and biological systems by producing informative structures associated with characteristic patterns in the sensory environment. In this work, we examined unsupervised latent representations of images of basic geometric shapes with neural network models of unsupervised generative self-learning. Background: Unsupervised concept learning with generative neural network models. Objective: Investigation of structure, geometry and topology in the latent representations of generative models that emerge as a result of unsupervised self-learning with minimization of generative error. Examine the capacity of generative models to abstract and generalize essential data characteristics, including the type of shape, size, contrast, position and orientation. Methods: Generative neural network models, direct visualization, density clustering, and probing and scanning of latent positions and regions. Results: Structural consistency of latent representations; geometrical and topological characteristics of latent representations examined and analysed with unsupervised methods. Development and verification of methods of unsupervised analysis of latent representations. Conclusion: Generative models can be instrumental in producing informative compact representations of complex sensory data correlated with characteristic patterns.

Absorbers are devices that internally consume electromagnetic waves to partially or completely attenuate them. The basic idea is to absorb electromagnetic radiation by resonating the intended surface with the incident electromagnetic waves. This article focuses on the development of the absorber (from single-band to multi-band, narrow to broadband, non-tunable to tunable, and so on). The basic absorption principle of the current popular and excellent metamaterial graphene absorber is provided, as is the theoretical explanation of impedance matching and how to attain critical performance metrics like tunability, as well as prospects for terahertz (THz) absorber applications. Finally, numerous innovative absorbers are shown as examples, providing new ideas for future researchers.

Background and Objectives: Patients with intractable cutaneous injuries and complications such as severe ulcers and dysfunctions often result in unfavorable prognosis and burdensome possessions esthetically and psychosocially. This study aimed to evaluate the preliminary safety and efficacy outcomes of hyaluronic acid/placenta-derived mesenchymal stem/stromal cells (HA/P-MSCs) composite upon extensive traumas. Methods: The phase I of the interventional study with four enrolled patients with refractory wounds and ulcerations was conducted. With the consent of the patients and approval of the ethics committee, continuous HA/P-MSC composite administration was conducted on the surface of wounds after surgical debridement. The dynamic changes in pathomorphology and therapeutic response were collectively recorded. Results: All the participants revealed efficacious remission with clinical-grade HA/P-MSC composite, including complete wound healing, elimination of inflammatory exudate and refractory ulcers, and regeneration of damaged tissues under sterile conditions. No untoward effects or recurrence were observed in the cases during treatment and the 24-month’s follow-up visit. Conclusion: Refractory wound management with HA/P-MSC composite was preliminarily safe and effective for the improvement of the outcomes of the patients. These findings provided new references for the treatment of the costliest and incurable diseases associated with cutaneous injuries.

Aims: This work reports a 405 nm photodetector based on the thermal evaporated CsPbBr3-CsPb2Br5 compound thin film. Background: The post-annealing process of the CsPbBr3-CsPb2Br5 compound thin film prepared by the thermal evaporation method has been improved in this work. To enhance the crystallization and photoresponse properties of the thin films, dimethyl sulfoxide (DMSO) steam was used in the postannealing process instead of using the previous way that increased the annealing temperature. Methods: The CsPbBr3-CsPb2Br5 compound powder was deposited directly on the surface glass substrate by thermal evaporation to form the CsPbBrCsPbBr3-CsPb2Br5 compound thin film. The thin films were post-annealed at 150°C for 15 min to crystallize. The DMSO liquid was dropped on the substrate; the liquid would then evaporate completely, leading to the formation of DMSO steam during 150°C post-annealing. The DMSO steam would cover the thin film completely, assisting with crystallization. Finally, the gold electrodes were deposited on the surface of thin films with a conductive channel of 1 mm´ 100 mm. Results: Results showed the crystalline quality of the thin film after DMSO steam annealing to be greatly improved compared to that of the thin film without DMSO steam annealing. The energy gap was between 2.355 eV and 2.293 eV, which was similar to a previous report. In addition, under 405 nm excitation, the photocurrent of the thin film annealed in DMSO steam showed a rapid response (35 ms), good dependence of photocurrent on light radiation power, and improved responsivity. Especially, the responsivity at 3 V bias of the thin film annealed in DMSO steam increased to 1.5 times that of the thin film without DMSO steam annealing and even 4.5 times that of as-deposited film. Conclusion: A 405 nm photodetector based on the thermal evaporated CsPbBr3-CsPb2Br5 compound thin film was prepared successfully. The newest report has shown improved preparation process of CsPbBr3-CsPb2Br5 compound thin films, where low annealing temperature with the DMSO steam-assisted post-annealing process was used. The thin film annealed in DMSO steam was found to possess high crystalline quality and enhanced photoresponse performance, compared to thin film without DMSO steam annealing.

Background: Hong Kong is situated on the coast of the south-eastern tip of China. A sub-tropical climate brings this coastal city a rainy season from April to October each year. Rainfall intensities as high as 50 mm to 100 mm per hour and 250 mm to 350 mm in 24 hours are not uncommon. With the hilly terrain and densely populated development, Hong Kong is susceptible to land-sliding during heavy rainstorms. Concerted efforts have been made by the Geotechnical Engineering Office (GEO) of the Hong Kong Special Administrative Region Government, to operate and enhance a territory-wide Landslip Warning System continuously, as a major component of landslide risk management tools in Hong Kong, since its establishment in 1977. Forewarning the public of possible landslide risks due to heavy rainstorms and triggering an emergency system within the government to deal with landslide incidents are the major objectives of the Landslip Warning System. This paper presents the development and enhancement of the comprehensive people-centred early warning system for landslides in Hong Kong, based on a framework of an early warning system with four major elements, including risk knowledge, monitoring and warning services, communication and dissemination, and response capability. Methods: With good quality landslide, rainfall and slope databases, landslide prediction models have been developed and continuously updated using statistical means to assess the severity of landslides based on real-time rainfall and rainfall forecast. For the purpose of an effective and robust operation of the Landslip Warning System, the latest Internet of Things (IoT) and cloud computing technologies have been deployed to operate the GEO Rainguage System collecting, transmitting and processing real-time rainfall data over the territory of Hong Kong, and to enhance system redundancy and resilience for data analysis and information dissemination. Results: To have an effective communication and dissemination of the Landslip Warning information to the community, various public education activities have been carried out regularly to promote public awareness of landslide risks and promote personal precautionary measures during heavy rainstorms. Conclusion: Well-established procedures for the operation of the Landslip Warning System have been developed to trigger the internal emergency mechanism within the government to deal with the landslide danger, with the support of innovative solutions to enhance the capability and efficiency of emergency responses.

The accurate description of strongly correlated systems, also known as multireference systems, requires a balanced treatment of static and dynamic correlations and is an important target for developing quantum chemical methods. An appealing treatment to economically describe strongly correlated systems is the multireference density function theory (MRDFT) approach, in which the static correlation is included in the multiconfigurational wave function, while the density functional includes the dynamic correlation. This mini-review focuses on the recent progress and applications of the density functional methods based on valence bond theory. A series of density functional valence bond (DFVB) methods are surveyed, including the dynamic correlation correction- based and Hamiltonian matrix correction-based DFVB methods, the hybrid one-parameter DFVB methods, the block-localized density functional theory and the multistate density functional theory. These methods have been applied to various chemical and physical property calculations of strongly correlated systems, including resonance energies, potential energy curves, spectroscopic constants, atomization energies, spin state energy gaps, excitation energies, and reaction barriers. Most of the test results show that the density functional methods based on VB theory give comparable accuracy but require lower computational cost than high-level quantum computational methods and thus provide a promising strategy for studying strongly correlated systems.

Background: The covariance matching procedure of the measurement noise covariance, namely the R matrix, was processed in singular value decomposition (SVD), which is one of the single-frame methods. Aims: Tuning the system noise covariance Q matrix for the single-frame method aided Kalman filtering algorithm. Objective: Develop the R and Q double covariance matching rule for the single-frame method aided Kalman filtering algorithm. Methods: The matching procedure of the measurement noise covariance, namely the R matrix, is processed in singular value decomposition (SVD), which is one of the single-frame methods. The second matching rule is defined in the second stage of the proposed EKF design. Results: The matching rules are run simultaneously, which makes the filter capable of being robust against initialization errors, system noise uncertainties, and measurement malfunctions at the same time without an external filter design necessity. Conclusion: A single-frame method aided Kalman filtering algorithm based double covariance matching rule is presented in this paper. First, the measurement noise covariance matching is introduced using the SVD method that processes the R-adaptation inherently for the filtering stage. Second, the system noise covariance matching is described so as to have double covariance matching at the same time during the estimation procedure. The SVD-Aided AEKF becomes R- and Qadaptive simultaneously by applying the Q-adaptation rule to the intrinsically R-adaptive SVDaided EKF.