Search

The dynamic risk assessment of typhoon disasters is an important decision-making basis for disaster response in the event of a major typhoon. Its goal is to dynamically predict the expected loss and disaster risk level caused by a typhoon so as to provide a basis for disaster risk early warning and emergency response. The traditional risk assessment model mainly fits the vulnerability curves of the hazard-affected bodies using historical disaster losses, and then establishes a disaster risk assessment model by coupling the risk of disaster factors, exposure, and vulnerability. However, the vulnerability curves generated by this method have problems of regional applicability, particularly in small-scale regions with small sample sizes available for fitting, leading to insufficient generalizability of the model. In addition, such models are complex and require phased hazard and vulnerability of the hazard-affected bodies research. Moreover, when employing the 3-element coupling process, it is difficult to consider other risk factors in the disaster system, such as hazard-formative environment and disaster prevention and mitigation capability. With the development of information technology, the availability of disaster risk factor data has been significantly improved, affording conditions for the fusion and application of disaster risk multi-source data. In recent years, many data-driven machine-learning models have been used to establish disaster risk assessment models. These models have the advantage that they can use large sample to improve the adaptability of the model, whereby the modeling process can consider more risk factors, the concepts of hazard and vulnerability are diluted, and the steps of model building are simplified. The integrated learning algorithm can not only improve the prediction accuracy, but more importantly, can be used to effectively evaluate the contribution value of the index to the final evaluation result. At present, China has established a six-level disaster reporting system at the national, provincial, municipal, county, township, and village levels, forming a long-term, high-precision database of disaster event cases since 2009, providing rich disaster loss information for the data fusion of risk elements. This study was based on 108 typhoon cases affecting five provinces in southeast China during 2009-2022. Nearly 4,000 county-level typhoon disaster loss samples were used to establish a dynamic typhoon risk assessment sample database that integrates 30 types of multi-source risk factor indicators. Six typhoon disaster risk assessment models were developed using the random forest algorithm to evaluate the affected population, emergency relocation population, crop-affected areas, collapsed and severely damaged houses, direct economic losses, and comprehensive risk level. Through the verification of actual disaster situations and model results, the overall accuracy of the disaster risk assessment results was found to be greater than 80%, indicating that the model has good generalizability and can be used for actual disaster assessment work. The experimental comparison shows that increasing the training sample size by 1-2 orders of magnitude can improve the accuracy of the model assessment by 3%-14%, indicating that the accumulation of disaster risk big data is of great significance in the study of disaster risk assessment. This study is expected to constitute a scientific reference for the quantitative analysis of the multiple impact factors of typhoon disaster damage and explore technical ideas for the application of disaster big data in risk management.

The proliferation of artificial intelligence models has rendered them vulnerable to a myriad of security threats. The extensive integration of deep learning models into edge devices has introduced novel security challenges. Given the analogous structural characteristics of deep neural networks, adversaries can employ decompilation tactics to extract model structural details and parameters, facilitating the reconstruction of these models. Such actions can compromise the intellectual property rights of the model and increase the risk of white-box attacks. To mitigate the capability of model decompilers to locate and identify model operators, acquire parameters, and parse network topologies, an obfuscation framework was proposed. This framework was embedded within the model compilation process to safeguard against model extraction attacks. During the frontend optimization phase of deep learning compilers, three obfuscation techniques were developed and integrated: operator obfuscation, parameter obfuscation, and network topology obfuscation. The framework introduced opaque predicates, incorporated fake control flows, and embedded redundant memory access to thwart the reverse engineering efforts of model decompilers. The experimental findings demonstrate that the obfuscation framework, named DNNobfus, significantly diminishes the accuracy of state-of-the-art model decompilation tools in identifying model operator types and network connections to 21.63% and 48.24%, respectively. Additionally, DNNobfus achieves an average time efficiency of 67.93% and an average space efficiency of 88.37%, surpassing the performance of the obfuscation tool Obfuscator-LLVM in both respects.

BackgroundViral tracers are important tools for mapping brain connectomes. The feature of predominant anterograde transneuronal transmission offers herpes simplex virus-1 (HSV-1) strain H129 (HSV1-H129) as a promising candidate to be developed as anterograde viral tracers. In our earlier studies, we developed H129-derived anterograde polysynaptic tracers and TK deficient (H129-dTK) monosynaptic tracers. However, their broad application is limited by some intrinsic drawbacks of the H129-dTK tracers, such as low labeling intensity due to TK deficiency and potential retrograde labeling caused by axon terminal invasion. The glycoprotein K (gK) of HSV-1 plays important roles in virus entry, egress, and virus-induced cell fusion. Its deficiency severely disables virus egress and spread, while only slightly limits viral genome replication and expression of viral proteins. Therefore, we created a novel H129-derived anterograde monosynaptic tracer (H129-dgK) by targeting gK, which overcomes the limitations of H129-dTK.MethodsUsing our established platform and pipeline for developing viral tracers, we generated a novel tracer by deleting the gK gene from the H129-G4. The gK-deleted virus (H129-dgK-G4) was reconstituted and propagated in the Vero cell expressing wildtype H129 gK (gKwt) or the mutant gK (gKmut, A40V, C82S, M223I, L224V, V309M), respectively. Then the obtained viral tracers of gKmut pseudotyped and gKwt coated H129-dgK-G4 were tested in vitro and in vivo to characterize their tracing properties.ResultsH129-dgK-G4 expresses high levels of fluorescent proteins, eliminating the requirement of immunostaining for imaging detection. Compared to the TK deficient monosynaptic tracer H129-dTK-G4, H129-dgK-G4 labeled neurons with 1.76-fold stronger fluorescence intensity, and visualized 2.00-fold more postsynaptic neurons in the downstream brain regions. gKmut pseudotyping leads to a 77% decrease in retrograde labeling by reducing axon terminal invasion, and thus dramatically improves the anterograde-specific tracing of H129-dgK-G4. In addition, assisted by the AAV helper trans-complementarily expressing gKwt, H129-dgK-G4 allows for mapping monosynaptic connections and quantifying the circuit connectivity difference in the Alzheimer's disease and control mouse brains.ConclusionsgKmut pseudotyped H129-dgK-G4, a novel anterograde monosynaptic tracer, overcomes the limitations of H129-dTK tracers, and demonstrates desirable features of strong labeling intensity, high tracing efficiency, and improved anterograde specificity.

This study evaluates aspects related to P12 principals' professional development needs in South Carolina regarding the three domains of school leadership: management, instructional leadership, and program administration. A survey to rate principals' current leadership knowledge, rank order their professional development needs, and provide a confidence rating regarding their abilities was given to over 1,100 principals and 85 superintendents. Through examining relationships with a psychometric model, results derived latent leadership ability scores and self-reported confidence ratings of principals as well as the superintendents' leadership scores and confidence ratings of their principals. This study found a significant discrepancy between principals' and superintendents' confidence ratings and their corresponding leadership ability scores, respectively. A further analysis of the rank-ordered professional development needs highlighted instructional leadership to be the most needed topic for professional development. Finally, atypical response patterns regarding principal's current leadership knowledge are also identified through person-fit analysis to provide additional information regarding P-12 principals' professional development needs.