Search

Backdoor attacks have been recognized as a major AI security threat in deep neural networks (DNNs) recently. The attackers inject backdoors into DNNs during the model training such as federated learning. The infected model behaves normally on the clean samples in AI applications while the backdoors are only activated by the predefined triggers and resulted in the specified results. Most of the existing defensing approaches assume that the trigger settings on different poisoned samples are visible and identical just like a white square in the corner of the image. Besides, the sample-specific triggers are always invisible and difficult to detect in DNNs, which also becomes a great challenge against the existing defensing protocols. In this paper, to address the above problems, we propose a backdoor detecting and mitigating protocol based on a wider separate-then-reunion network (WISERNet) equipped with a cryptographic deep steganalyzer for color images, which detects the backdoors hiding behind the poisoned samples even if the embedding algorithm is unknown and further feeds the poisoned samples into the infected model for backdoor unlearning and mitigation. The experimental results show that our work performs better in the backdoor defensing effect compared to state-of-the-art backdoor defensing methods such as fine-pruning and ABL against three typical backdoor attacks. Our protocol reduces the attack success rate close to 0% on the test data and slightly decreases the classification accuracy on the clean samples within 3%.

A polyether, acrylic acid and oleic acid were used as raw materials to synthesise oleic acid modified polycarboxylate superplasticiser (OA-PCE) by emulsion polymerisation. The synthesis process was optimised using orthogonal experiments. The colloidal interfacial chemistry results showed that the OA-PCE was more capable of reducing the surface tension of aqueous solutions than an unmodified traditional polycarboxylate superplasticiser (JS-PCE) and its adsorption capacity on cement particles was greater. The isothermal adsorption behaviour of the two kinds of PCEs on cement conformed to the Freundlich isothermal adsorption equation; it was a physical adsorption process by electrostatic action. The effect of OA-PCE dosage on the flowability of cement slurry was preliminarily investigated by rheological behaviour tests. According to the yield stress, plastic viscosity and thixotropic ring area of the slurry, a dosage of 0.35 wt% OA-PCE produced the optimum effect. The compressive strength of concrete made with 0.35 wt% OA-PCE at 4 h and 7 days was, respectively, 117.1% and 88.3% higher than that of the control group and met the standards for the use of concrete in metro pipe sheets. OA-PCE synthesised by emulsion polymerisation is therefore significant for the development of PCEs both in terms of the polymerisation method and practical applications.

As an important component of corporate inequality, stock market concentration has become a focus of attention in academia in recent years. However, existing literature focuses on its negative consequences, and research on the determinants of stock market concentration is scarce. This paper investigates for the first time how stock market downturns affect stock market concentration. Using data on stock markets in both the United States and China, we find a negative correlation between market-wide returns and stock market concentration. To address endogeneity and establish causal inference, we exploit two natural experiments: the COVID-19 pandemic and the subprime crisis. We find that stock market concentration increases during these crises, and we also find some heterogeneity between the United States and China. Our findings have important policy implications regarding inequality during market downturns.

This article has conducted relevant surveys on the current situation of energy use in rural areas of Linzhi City, Tibet, and Heishui County, Sichuan. The survey results show that in terms of energy sources, fuelwood accounts for the largest proportion in rural areas of Linzhi City, Tibet, accounting for 72.2% to 74.7% of energy use. Moreover, the proportion of renewable energy in rural areas is 4.5 times that of pastoral areas, and 18 times that of resource rich areas. In rural areas of Heishui County, Sichuan Province, dried cow and sheep dung is the main energy source, accounting for 79.8% of the total energy consumption. The utilization rate of renewable energy in daily life is relatively low, only 0.3%.

The dynamics of gene expression in crop grains has typically been investigated at the transcriptional level. However, this approach neglects translational regulation, a widespread mechanism that rapidly modulates gene expression to increase the plasticity of organisms. Here, we performed ribosome profiling and polysome profiling to obtain a comprehensive translatome data set of developing bread wheat (Triticum aestivum) grains. We further investigated the genome-wide translational dynamics during grain development, revealing that the translation of many functional genes is modulated in a stage-specific manner. The unbalanced translation between subgenomes is pervasive, which increases the expression flexibility of allohexaploid wheat. In addition, we uncovered widespread previously unannotated translation events, including upstream open reading frames (uORFs), downstream open reading frames (dORFs), and open reading frames (ORFs) in long noncoding RNAs, and characterized the temporal expression dynamics of small ORFs. We demonstrated that uORFs act as cis-regulatory elements that can repress or even enhance the translation of mRNAs. Gene translation may be combinatorially modulated by uORFs, dORFs, and microRNAs. In summary, our study presents a translatomic resource that provides a comprehensive and detailed overview of the translational regulation in developing bread wheat grains. This resource will facilitate future crop improvements for optimal yield and quality.

The purpose of this study is to explore the effect of using 3D printed pelvic prosthesis to reconstruct bone defect after pelvic tumor resection. From June 2018 to October 2021, a total of 10 patients with pelvic tumors underwent pelvic tumor resection and 3D printed customized hemipelvic prosthesis reconstruction in our hospital. Enneking pelvic surgery subdivision method was used to determine the degree of tumor invasion and the site of prosthesis reconstruction. 2 cases in Zone I, 2 cases in Zone II, 3 cases in Zone I + II, 2 cases in Zone II + III and 1 case in Zone I + II + III. Patients had preoperative VAS scores of 6.5 ± 1.3, postoperative VAS scores of 2.2 ± 0.9, preoperative MSTS-93 scores of 9.4 ± 5.3 and postoperative 19.4 ± 5.9( p < 0.05), all patients had improvement in pain after surgery; Postoperative complications included joint dislocation in 2 cases, myasthenia caused by Guillain-Barre syndrome in 1 case, delayed wound healing in 3 cases and wound infection in 2 cases. Postoperative wound-related complications and dislocations were associated with the extent of the tumor. Patients with tumor invasion of the iliopsoas and gluteus medius muscles had higher complication rates and worse postoperative MSTS scores ( p < 0.05). The patients were followed up for 8 ∼ 28 months. During the follow-up period, 1 case recurred, 4 cases metastasized and 1 case died. All pelvic CTs reviewed 3–6 months after surgery showed good alignment between the 3D printed prosthesis and the bone contact, and tomography showed the growth of trabecular structures into the bone. Overall pain scores decreased and functional scores improved in patients after 3D printed prosthesis replacement for pelvic tumor resection. Long-term bone ingrowth could be seen on the prosthesis-bone contact surface with good stability.

Due to the diversity of the degradation process that is difficult to model, the recovery of mixed distorted images is still a challenging problem. The deep learning model trained under certain degradation declines significantly in other degradation situations. In this article, we explore ways to use a combination of tools to deal with the mixed distortion. First, we illustrate the limitations of a single deep network in dealing with multiple distortion types and then introduce a hierarchical toolkit with distinguished powerful tools. Second, we investigate how an efficient representation of images combined with a reinforcement learning (RL) paradigm helps to deal with tool noise in continuous restoration. The proposed method can accurately capture the distortion preferences for selecting the optimal recovery tools by RL agent. Finally, to fully utilize random tools for unknown distortion combinations, we adopt the exploration scheme with various quality evaluation methods to achieve more quality improvements. Experimental results demonstrate that the peak signal-to-noise ratio of the proposed method is 3.30 dB higher than other state-of-the-art RL-based methods on the CSIQ single distortion dataset and 0.95 dB higher on the DIV2K mixed distortion dataset.

The delayed platelet engraftment associated with allogeneic hematopoietic stem cell transplantation (allo-HSCT) is a common complication and often results in increased transplant-related complications. A single-center, prospective, investigator-initiated pilot study was conducted to explore whether herombopag, a 2nd generation thrombopoietin-receptor agonist, would promote platelet engraftment after allo-HSCT. Between 2/2022 and 06/2022, seventeen individuals (median age 39; range 15-58 years) with hematological malignancies were enrolled. Herombopag was given for a median of 22 (range 14-61) days at a dose of 7.5 mg/d. The median time to neutrophil500/μl was 11 (range 9-19) days. The median time to platelet20,000/μl and50,000/μl was 13 (range 8-22), and 20 (range 14-45) days, respectively. Compared with historical controls, the cumulative incidence of platelet engraftment after HSCT was significantly higher in the herombopag group (20,000/μl at day +21, 88% vs 65%, p=0.003;50,000/μl at day +30, 65% vs 43%, p=0.001). Herombopag also reduced the units of platelet transfusion within 30 days post-SCT (3.6±2.5 vs 5.4±3.2 U, p=0.01). In conclusion, it seems likely that herombopag could enhance platelet engraftment after allo-HSCT.

Objectives. The aim of the study is to evaluate the efficacy and prognosis of neoadjuvant chemotherapy (NAC) combined with breast-conserving surgery (BCS) in treating triple-negative breast cancer (TNBC) and analyze the influencing factors and predictors of the efficiency and prognosis of NAC. Methods. A retrospective cohort study was conducted by dividing patients into two groups according to two different therapy methods. With BCS as the exposure factor, 46 cases were assigned to the exposed group and 80 cases to the nonexposed group. We compare the difference in operation-related indicators, postoperative complications, local recurrence rate, distant metastasis rate, and overall survival (OS) rate between the two groups. The factors affecting the efficiency and prognosis of NAC were analyzed by binary logistic regression, and the optimal cutoff value was determined by the area under the ROC curve (AUC). The survival curve was plotted, and the univariate log-rank test was performed to analyze the difference in OS between the two groups. The influencing factors of OS were analyzed by the Cox risk regression model. Results. NAC + BCS resulted in significantly less intraoperative blood loss, lower incidence of postoperative complications, and shorter operative time and length of hospital stay than that in NAC ( P < 0.05 ). There was no significant difference in local recurrence, distant metastasis, or OS between the two groups ( P > 0.05 ). Multivariate analysis showed that the clinical stage I and Ki-67 high expression were independent protective factors of the efficacy of NAC. The high expression of Ki-67 and nondecline expression of Ki-67 were independent risk factors of prognosis. Ki-67 high expression was an independent risk factor of OS ( P < 0.05 ). The ROC curve showed that the AUC of Ki-67 for NAC efficacy, prognosis, and OS were 0.706, 0.820, and 0.687, respectively, with optimal cutoff values of 25.5%, 29.0%, and 32.5%, respectively. Survival analysis showed that the OS of patients with NAC + BCS was 73.9% and NAC + MRM was 70.0% ( P > 0.05 ). In the low expression subgroup of Ki-67, the OS of the two groups were 100.0% and 77.8%, respectively ( P = 0.060 ). In the high expression subgroup of Ki-67, the OS of the two groups were 53.8% and 63.6%, respectively ( P = 0.419 ). Conclusions. NAC + BCS is a good method for treating TNBC, which has an obvious short-term effect and a good long-term prognosis. Clinical stage I and the high expression of Ki-67 are independent protective factors for the efficacy of NAC. The high expression of Ki-67 and nondecline expression of Ki-67 are independent risk factors of prognosis. Ki-67 is a potential predictor for the efficacy, prognosis, and OS of NAC in TNBC patients. The high expression of Ki-67 indicates better NAC efficacy, a poorer prognosis, and a lower OS.

As global economic integration advances, an increasing number of businesses are confronted with the conflict between economic and social benefits, as well as between the pursuit of corporate profits and the observance of ethics and morals. Using the Employee Work Ethics Questionnaire, the Organizational Commitment Questionnaire, and the Organizational Citizenship Behavior Questionnaire, this study focuses on the differences between different generations of employees' work ethics and organisational commitment and organisational citizenship behaviors. The study focuses on the enterprise management of "post-80s, post-90s, and post-00s" personnel and explores the distinctions between "post-80s," "post-90s," and "post-00s" people. The objective is to assist the "post-80s," "post-90s," and "post-00s" in adapting to corporate management with related measures. The objective is to assist the "post-80s," "post-90s," and "post-00s" in better adapting to enterprise management in order to achieve the enterprise's sustainable development. Several company personnel were chosen as responders, and the data are analyzed using Questionnaire Star and SPSS. A regression model of employee work ethics, organisational commitment, and organisational citizenship behavior was established. Generational differences were discovered in the effects of gender, years of work experience, education, and position on organisational commitment and organisational citizenship behaviors. However, in all three age groups, work ethics was strongly and positively associated with organisational commitment and organisational citizenship behaviors.