Search

Although deep learning models show powerful performance, they are still easily deceived by adversarial samples. Some methods for generating adversarial samples have the drawback of high time loss, which is problematic for adversarial training, and the existing adversarial training methods are difficult to adapt to the dynamic nature of the model, so it is still challenging to study an efficient adversarial training method. In this paper, we propose an adversarial training method, the core of which is the improved adversarial sample generation method AGFAT for adversarial training and the improved dynamic adversarial training method AGFAT-DAT. AGFAT uses a word frequency-based approach to identify significant words, filter replacement candidates, and use an efficient semantic constraint module as a means to reduce the time of adversarial sample generation; AGFAT-DAT is a dynamic adversarial training approach that uses a cyclic attack on the model after adversarial training and generates adversarial samples for adversarial training again. It is demonstrated that the proposed method can significantly reduce the generation time of adversarial samples, and the adversarial-trained model can also effectively defend against other types of word-level adversarial attacks. [ABSTRACT FROM AUTHOR]

The 14-3-3 proteins are highly conserved molecules that are involved in many vital biologic processes and are associated with the progression of cancer. The role of 14-3-3ζ, a dimeric isoform of 14-3-3, in intrahepatic cholangiocarcinoma (ICC) was investigated in this study. The expression of 14-3-3ζ in tumour samples from patients with ICC was examined by Western blot and immunohistochemistry, and the correlation between its expression and various clinicopathological features was determined. Then, the capacity for invasion, migration and proliferation as well as the expression of epithelial-mesenchymal transition (EMT)-related markers in ICC cells were assessed after 14-3-3ζ depletion. Finally, the prognostic significance of 14-3-3ζ in patients with ICC was further evaluated by Kaplan-Meier and Cox regression analyses. The expression of 14-3-3ζ was significantly higher in ICC tissues compared to peritumoural tissues. High expression of 14-3-3ζ positively correlated with lymphatic metastasis and tumour-node-metastasis (TNM) stage. The inhibition of 14-3-3ζ expression was able to impair the invasion, migration and proliferation of ICC cells in vitro. The expression of 14-3-3ζ was significantly correlated with the expression of the EMT-related markers snail and E-cadherin in ICC samples. Moreover, the down-regulation of 14-3-3ζ also decreased the phosphorylation of extracellular signal-regulated kinase (ERK) in ICC cells. Clinically, patients with ICC with high 14-3-3ζ expression demonstrated a poor prognosis in terms of a short overall survival and a high recurrence rate of the disease. A multivariate analysis revealed that 14-3-3ζ overexpression was an independent prognostic indicator for patients with ICC. 14-3-3ζ may enhance the invasive and proliferative capacity of tumour cells and thus prompt the progression of ICC via the activation of ERK signalling and the induction of EMT. The overexpression of 14-3-3ζ may be used as a prognostic biomarker and therapeutic target in patients with ICC.

Using the static opaque chamber method, the soil respiration rates (SR) were measured through the continuous experiments in situ in semiarid Stipa grandis steppe in Xilin River Basin of Inner Mongolia, China from June 2001 to June 2003, in parallel, the difference between the SR and the ecosystem respiration rates (TER) were compared. The results indicated that the seasonal variations of the SR and TER were obvious with higher emissions in growing season and a relatively low efflux level in non-growing season, furthermore, the negative effluxes were found in the observation site in winter; the annual CO2 efflux of total ecosystem ranged from 160.5 gC/(m2 x a) to 162.8 gC/(m2 x a) and that of soil ranged from 118.7 gC/(m2 x a) to 152.3 gC/(m2 x a). The annual SR accounted for about 74.0% to 93.5% of the annual TER, but the results of Analysis of Variance (ANOVA) indicated that the difference between the annual average TER and SR did not reach the significance level of 0.05. The TER was under similar environmental controls as SR, in growing seasons of drought years, the variations of soil moisture at 0-10 cm and 10-20 cm depth could account for 79.1%-95.6% of the changes of the SR and TER, but in non-growing season, more than 75% of the variations of the SR and TER could be explained by the changes of the ground temperature of soil surface layers.