While deep neural networks have achieved great success in graph analysis, recent work has shown that they are vulnerable to adversarial attacks. Compared with adversarial attacks on image classification, performing adversarial attacks on graphs is more challenging because of the discrete and non-differential nature of the adjacent matrix for a graph. In this work, we propose Cluster Attack -- a Graph Injection Attack (GIA) on node classification, which injects fake nodes into the original graph to degenerate the performance of graph neural networks (GNNs) on certain victim nodes while affecting the other nodes as little as possible. We demonstrate that a GIA problem can be equivalently formulated as a graph clustering problem; thus, the discrete optimization problem of the adjacency matrix can be solved in the context of graph clustering. In particular, we propose to measure the similarity between victim nodes by a metric of Adversarial Vulnerability, which is related to how the victim nodes will be affected by the injected fake node, and to cluster the victim nodes accordingly. Our attack is performed in a practical and unnoticeable query-based black-box manner with only a few nodes on the graphs that can be accessed. Theoretical analysis and extensive experiments demonstrate the effectiveness of our method by fooling the node classifiers with only a small number of queries.
With the rapid development of information technology, online education has become a new teaching mode in colleges and universities, and has made great contributions to China’s education reform. The online education of ideological and political education for college students is still in its infancy. With the deepening and popularization of the mobile Internet, the use of mobile ends to carry out online ideological and political education courses is becoming a mainstream choice. However, the large number of click-to-play demands caused the problem of response delay. In order to solve this problem, this paper proposes an online ideological and political education architecture based on edge computing. In order to effectively utilize the edge computing architecture and optimize the transmission delay, this paper introduces a caching approach for the Genetic Algorithm utilized in Long and Short-term Memory Network, referred to as LSTM-GA. First, in the face of the fast refresh of video requests when user ends move between different edge computing servers, based on the "device-edge-cloud" system collaborative architecture, to reduce transmission delay and optimize QoE to build a network model for the target. Furthermore, the genetic algorithm optimized Long-Short-Term Memory Network (LSTM-GA) model is applied to forecast the trajectory of the user’s endpoint, and seek the optimal solution with the minimum transmission delay, so as to ensure that the video content is cached in an appropriate location. According to the simulation results, the technique presented in this paper can significantly decrease transmission delays and enhance the quality of user experience.
This paper introduces a novel problem, distributional information embedding, motivated by the practical demands of multi-bit watermarking for large language models (LLMs). Unlike traditional information embedding, which embeds information into a pre-existing host signal, LLM watermarking actively controls the text generation process--adjusting the token distribution--to embed a detectable signal. We develop an information-theoretic framework to analyze this distributional information embedding problem, characterizing the fundamental trade-offs among three critical performance metrics: text quality, detectability, and information rate. In the asymptotic regime, we demonstrate that the maximum achievable rate with vanishing error corresponds to the entropy of the LLM's output distribution and increases with higher allowable distortion. We also characterize the optimal watermarking scheme to achieve this rate. Extending the analysis to the finite-token case, we identify schemes that maximize detection probability while adhering to constraints on false alarm and distortion.
Retroperitoneal fibrosis (RPF) is an uncommon condition characterized by inflammation and fibrosis in the retroperitoneal space. More than two-thirds of RPF are idiopathic, with the remaining stemed from a variety of secondary causes. It was suggested that IgG4-related RPF is a secondary form of RPF. We undertook this study to compare detailed demographic, clinical and laboratory characteristics of IgG4-related RPF and IRPF in a large Chinese cohort. We retrospectively reviewed the medical records of 132 RPF patients diagnosed at Peking University People’s Hospital between March 2010 and March 2018. Among the 132 patients, the mean age at disease onset was 54.8 years. IgG4-related RPF group showed greater male predominance compared to IRPF group. IgG4-related RPF patients showed a longer interval between symptom onset and diagnosis, and allergic diseases were more common in this group. Sixty-four patients (48.4%) had lower back pain, which was more common in IRPF group than that in IgG4-related RPF patients. In terms of organ involvement, although 42 of 47 patients (89.3%) with IgG4-related RPF had other organ involvement, there were no patients in the IRPF group with other organ involvement. In addition, the serum IgG4 level, elevated eosinophils counts and IgE level were significantly higher in IgG4-related RPF patients. We described the demographic, clinical and laboratory differences between IgG4-related RPF and IRPF patients, indicating their potential differences in pathogenesis, which was of great importance to diagnose and manage the two phenotypes.
Family-based studies provide a unique opportunity to characterize genetic risks of diseases in the presence of population structure, assortative mating, and indirect genetic effects. We propose a novel framework, PGS-TRI, for the analysis of polygenic scores (PGS) in case-parent trio studies for estimation of the risk of an index condition associated with direct effects of inherited PGS, indirect effects of parental PGS, and gene-environment interactions. Extensive simulation studies demonstrate the robustness of PGS-TRI in the presence of complex population structure and assortative mating compared to alternative methods. We apply PGS-TRI to multi-ancestry trio studies of autism spectrum disorders (N
<p>Supplementary Figure S3: Natural logarithm hazard ratios of breast cancer for P-spline term scaled 313-SNP PRS with 4 degrees of freedom estimated by fully-adjusted Cox proportional hazards model in ARIC (1990-2015)</p>
'/%3e%3cuse xlink:href='%23a' transform='rotate(23.036 .093 25.536)'/%3e%3cuse xlink:href='%23a' transform='rotate(45.87 1.273 16.18)'/%3e%3cuse xlink:href='%23a' transform='rotate(69.945 .996 12.078)'/%3e%3cuse xlink:href='%23a' transform='rotate(20.66 -19.689 31.932)'/%3e%3c/svg%3e)