No abstract

Online continual learning for image classification studies the problem of learning to classify images from an online stream of data and tasks, where tasks may include different classes (class incremental) or data nonstationarity (domain incremental). One of the key challenges of continual learning is to avoid catastrophic forgetting (CF), i.e., forgetting old tasks in the presence of more recent tasks. Over the past few years, a large range of methods and tricks have been introduced to address the continual learning problem, but many have not been fairly and systematically compared under a variety of realistic and practical settings.To better understand the relative advantages of various approaches and the settings where they work best, this survey aims to (1) compare stateof-the-art methods such as Maximally Interfered Retrieval (MIR), iCARL, and GDumb (a very strong baseline) and determine which works best at different memory and data settings as well as better understand the key source of CF; (2) determine if the best online class incremental methods are also competitive in domain incremental setting; and (3) evaluate the performance of 7 simple but effective trick such as "review" trick and nearest class mean (NCM) classifier to assess their relative impact. Regarding (1), we

Online class-incremental continual learning (CL) studies the problem of learning new classes continually from an online non-stationary data stream, intending to adapt to new data while mitigating catastrophic forgetting. While memory replay has shown promising results, the recency bias in online learning caused by the commonly used Softmax classifier remains an unsolved challenge. Although the Nearest-Class-Mean (NCM) classifier is significantly undervalued in the CL community, we demonstrate that it is a simple yet effective substitute for the Softmax classifier. It addresses the recency bias and avoids structural changes in the fully-connected layer for new classes. Moreover, we observe considerable and consistent performance gains when replacing the Softmax classifier with the NCM classifier for several state-of-the-art replay methods.To leverage the NCM classifier more effectively, data embeddings belonging to the same class should be clustered and well-separated from those with a different class label. To this end, we contribute Supervised Contrastive Replay (SCR), which explicitly encourages samples from the same class to cluster tightly in embedding space while pushing those of different classes further apart during replay-based training. Overall, we observe that our proposed SCR substantially reduces catastrophic forgetting and outperforms state-of-the-art CL methods by a significant margin on a variety of datasets.

The consumption of pharmaceuticals and personal care products (PPCPs) for controlling and preventing the COVID-19 would have sharply increased during the pandemic. To evaluate their post-pandemic environmental impacts, five categories of drugs were detected in lakes and WWTP-river-estuary system near hospitals of Jinyintan, Huoshenshan and Leishenshan in the three regions (J, H and L) (Regions J, H and L) in Wuhan, China. The total amount of PPCPs (ranging from 2.61 to 1122 ng/L in water and 0.11 to 164 ng/g dry weight in sediments) were comparable to historical reports in Yangtze River basin, whereas the detection frequency and concentrations of ribavirin and azithromycin were higher than those of historical studies. The distribution of concerned drugs varied with space, season, media and water types: sampling sites located at WWTPs-river-estuary system around two hospitals (Regions L and J) usually had relatively high waterborne contamination levels, most of which declined in autumn; lakes had relatively low waterborne contamination levels in summer but increased in autumn. The potential risks of detected PPCPs were further evaluated using the multiple-level ecological risk assessment (MLERA): sulfamethoxazole and azithromycin were found to pose potential risks to aquatic organisms according to a semi-probabilistic approach and classified as priority pollutants based on an optimized risk assessment. In general, the COVID-19 pandemic did not cause serious pollution in lakes and WWTPs-river-estuary system in Wuhan City. However, the increased occurrence of certain drugs and their potential ecological risks need further attention. A strict source control policy and an advanced monitoring and risk warning system for emergency response and long-term risk control of PPCPs is urgent.

BackgroundCanine distemper, caused by Canine distemper virus (CDV), is a highly contagious and fatal systemic disease in free-living and captive carnivores worldwide. Recombinase polymerase amplification (RPA), as an isothermal gene amplification technique, has been explored for the molecular detection of diverse pathogens.MethodsA real-time reverse transcription RPA (RT-RPA) assay for the detection of canine distemper virus (CDV) using primers and exo probe targeting the CDV nucleocapsid protein gene was developed. A series of other viruses were tested by the RT-RPA.Thirty-two field samples were further tested by RT-RPA, and the resuts were compared with those obtained by the real-time RT-PCR.ResultsThe RT-RPA assay was performed successfully at 40 °C, and the results were obtained within 3 min–12 min. The assay could detect CDV, but did not show cross-detection of canine parvovirus-2 (CPV-2), canine coronavirus (CCoV), canine parainfluenza virus (CPIV), pseudorabies virus (PRV) or Newcastle disease virus (NDV), demonstrating high specificity. The analytical sensitivity of RT-RPA was 31.8 copies in vitro transcribed CDV RNA, which is 10 times lower than the real-time RT-PCR. The assay performance was validated by testing 32 field samples and compared to real-time RT-PCR. The results indicated an excellent correlation between RT-RPA and a reference real-time RT-PCR method. Both assays provided the same results, and R2 value of the positive results was 0.947.ConclusionsThe results demonstrated that the RT-RPA assay offers an alternative tool for simple, rapid, and reliable detection of CDV both in the laboratory and point-of-care facility, especially in the resource-limited settings.

Bis­(2-ethylhexyl)-2,3,4,5-tetrabromophthalate (TBPH), a novel brominated flame retardant, can potentially cause lipid metabolism disorder; however, its biological effects on lipid homeostasis remain unknown. We investigated its ability to cause nonalcoholic fatty liver disease (NAFLD) in zebrafish. Female zebrafish were fed a high-fat diet (HFD, 24% crude fat) or normal diet (ND, 6% crude fat), and exposed to TBPH (0.02, 2.0 μM) for 2 weeks. Consequently, HFD-fed fish showed a higher measured concentration of TBPH than ND-fed fish. Further, TBPH-treated fish in the HFD group showed higher hepatic triglyceride levels and steatosis. In comparison to ND-fed fish, treating HFD-fed fish with TBPH led to an increase in the concentration of several proinflammatory markers (e.g., TNF-α, IL-6); TBPH exposure also caused oxidative stress. In addition, the mRNA levels of genes encoding peroxisome proliferator-activated receptors were increased, and the transcription of genes involved in lipid synthesis, transport, and oxidation was upregulated in both ND- and HFD-fed fish. Both the ND and HFD groups also showed demethylation of the peroxisome proliferator-activated receptor-γ coactivator 1-α gene promoter, accompanied by the upregulation of tet1 and tet2 transcription. To summarize, we found that TBPH amplified the disruption of lipid homeostasis in zebrafish, leading to the enhancement of diet-induced NAFLD progression.

No abstract

Background/Aims: Circadian locomotor output cycles protein kaput (CLOCK) plays a key role in maintaining circadian rhythms and activation of downstream elements. However, its function on human female reproductive system remains unknown. Methods: To investigate the potential role of CLOCK, CLOCK-shRNAs were transfected into mouse 129 ES cells or injected into the ovaries of adult female mice. Western blotting was utilized to analyze the protein interactions and flow cytometry was used to assess apoptosis. Results: The expression of CLOCK peaked at the 6th week in the healthy fetuses. However, an abnormal expression of CLOCK was detected in fetuses from spontaneous miscarriage. To determine the effect of CLOCK on female fertility, a small hairpin RNA (shRNA) strategy was used to specifically knockdown the CLOCK gene expression in vitro and in vivo. Knockdown of CLOCK induced apoptosis in mouse embryonic stem (mES) cells and inhibited the proliferation in mES cells in vitro. CLOCK knockdown also led to decreased release of oocytes and smaller litter size compared with control in vivo. Conclusions: Collectively, theses findings indicate that CLOCK plays an important role in fertility and that the CLOCK knockdown leads to reduction in reproduction and increased miscarriage risk.