Search

Can we train a hybrid discriminative-generative model within a single network? This question has recently been answered in the affirmative, introducing the field of Joint Energy-based Model (JEM), which achieves high classification accuracy and image generation quality simultaneously. Despite recent advances, there remain two performance gaps: the accuracy gap to the standard softmax classifier, and the generation quality gap to state-of-the-art generative models. In this paper, we introduce a variety of training techniques to bridge the accuracy gap and the generation quality gap of JEM. 1) We incorporate a recently proposed sharpness-aware minimization (SAM) framework to train JEM, which promotes the energy landscape smoothness and the generalizability of JEM. 2) We exclude data augmentation from the maximum likelihood estimate pipeline of JEM, and mitigate the negative impact of data augmentation to image generation quality. Extensive experiments on multiple datasets demonstrate that our SADA-JEM achieves state-of-the-art performances and outperforms JEM in image classification, image generation, calibration, out-of-distribution detection and adversarial robustness by a notable margin. Our code is available at https://github.com/sndnyang/SADAJEM., Comment: Accepted to CVPR 2023

Osteosarcoma (OS) is an uncommon tumour that mainly affects bone in children and adolescents. The current treatment options of OS are of limited significance due to their immense side effects. In the present manuscript, we have developed a novel series of 1,2,3-triazole chalcone derivatives as potential agents against OS. The compounds were synthesized and evaluated for their PI3K and mTOR inhibitory activity using luminescent kinase assay, and Lance ultra assay, resp. The entire set of molecules showed significant to moderate inhibition of both kinases in the nanomolar range. The three most active compounds: 4e (N-(4-(3-(1-(4-bromophenyl)-1H-1,2,3-triazol-4-yl)acryloyl)phenyl)-4-nitrobenzamide), 4f (N-(4-(3-(1-(4-bromophenyl)-1H-1,2,3-triazol-4-yl)acryloyl)phenyl)-4-chlorobenzamide) and 4g (4-bromo-N-(4-(3-(1-(4-bromophenyl)-1H-1,2,3-triazol-4-yl)acryloyl)phenyl)benzamide), were evaluated for anti-cancer activity against human OScancer cell line (MG-63), liver cancer cell line (HepG2), lung cancer cell line (A549) and cervical cancer (HeLa), using MTT assay. Among the tested series, compound 4e showed a better inhibitory profile than gedatolisib against PI3K and was approximately comparable to that of gedatolisib against mTOR. The most significant inhibitory activity was observed for compound 4e against all cell lines (MG-63, HepG2, A549 and HeLa), still somewhat lower to comparable to that of gedatolisib, but with the highest potency against MG-63 cells. Compound 4e was further tested for anticancer activity against other OS cells and showed to be equipotent to gedatolisib against U2OS and Saos-2 cells. Moreover, it was also found non-toxic to normal cells (BEAS-2B and MCF 10A). The effect of compound 4e was further determined on apoptosis of Saos-2 cells by Annexin-PI assay, where it significantly amplified the percentage of apoptotic cells. Our study demonstrates the development of novel 1,2,3-triazole chalcone derivatives as potential agents against OS.

Background: In early 2020, non-pharmaceutical interventions (NPIs) were implemented in China to reduce and contain the coronavirus disease 2019 (COVID-19) transmission. These NPIs might have also reduced the incidence of hand, foot, and mouth disease (HFMD). Methods: The weekly numbers of HFMD cases and meteorological factors in 31 provincial capital cities and municipalities in mainland China were obtained from Chinese Center for Disease Control and Prevention (CCDC) and National Meteorological Information Center of China from 2016 to 2020. The NPI data were collected from local CDCs. The incidence rate ratios (IRRs) were calculated for the entire year of 2020, and for January–July 2020 and August–December 2020. The expected case numbers were estimated using seasonal autoregressive integrated moving average models. The relationships between kindergarten closures and incidence of HFMD were quantified using a generalized additive model. The estimated associations from all cities were pooled using a multivariate meta-regression model. Findings: Stringent NPIs were widely implemented for COVID-19 control from January to July 2020, and the IRRs for HFMD were less than 1 in all 31 cities, and less than 0·1 for 23 cities. Overall, the proportion of HFMD cases reduced by 52·9% (95% CI: 49·3−55·5%) after the implementation of kindergarten closures in 2020, and this effect was generally consistent across subgroups. Interpretation: The decrease in HFMD incidence was strongly associated with the NPIs for COVID-19. HFMD epidemic peaks were either absent or delayed, and the final epidemic size was reduced. Kindergarten closure is an intervention to prevent HFMD outbreaks. Funding: This research was supported by the National Natural Science Foundation of China (81973102 & 81773487), Public Health Talents Training Program of Shanghai Municipality (GWV-10.2-XD21), the Shanghai New Three-year Action Plan for Public Health (GWV-10.1-XK16), the Major Project of Scientific and T

Background: Hand, foot, and mouth disease (HFMD) is an important public health problem. A monovalent EV-A71 vaccine was launched in China in 2016. Previous studies showed that inactivated monovalent EV-A71 vaccines were highly efficient against HFMD associated with EV-A71 but not against HFMD with other etiologies, leading to a hypothesis that the introduction of EV-A71 vaccines might change the pathogen spectrum and epidemiological trend of HFMD. In this study, we described for the first time the changing epidemiological characteristics of HFMD after the launch of the EV-A71 vaccine. Methods: We extracted individual-based epidemiological data on HFMD cases reported to the Chinese Center for Disease Control and Prevention between January 2013 and December 2019. We described the changing epidemiological characteristics of HFMD before and after vaccine launch according to the distribution of diseases characteristics (demographic, temporal, and geographical) and evaluated the potential changes in risk factors of severe patients. All analyses were stratified by the phase before and after vaccine launch, and by enterovirus serotype. Findings: During 2013−2019, 15,316,710 probable cases of HFMD were reported. Of these, 787,197 (5·1%) were laboratory confirmed and 76,982 (0·5%) were severe. After the launch of the EV-A71 vaccine, the median age of HFMD patients infected with EV-A71 increased from 2·24 years (IQR:1·43, 3·56) to 2·81 years (IQR:1·58, 4·01). The proportion of patients less than 3 years of age decreased while the proportion of patients 3−5 years of age increased. There was a large decrease (60·7%) in the proportion of severe cases as well as a decline (28·3%) in HFMD patients infected with EV-A71. After the launch of the EV-A71 vaccine, the severe illness rate and mortality rate of HFMD patients in all age groups has decreased sharply, 62·20% and 83·78% respectively. The timing of the HFMD epidemic peak was delayed (1-2 months). After the launch of EV-A71 v