Search

The performance of deep neural networks is typically associated with their depth and width. As the performance of a model improves, it demands greater computational resources and memory. However, self-distillation offers a pathway to augment this performance by internally transferring knowledge. Traditional self-distillation methods lack guidance from teacher networks, thus lacking the ability to obtain correct and key information. In this paper, we propose a new self-distillation framework—Global key knowledge distillation framework (GK-KD). We implement error correction mechanisms to network logit outputs and assess the importance of feature maps through rank to obtain correct and important knowledge. Experiments demonstrate the effectiveness of this framework: on the CIFAR100 dataset, ResNet18 achieved a 4.56% accuracy improvement over the baseline, ResNet152 also achieved a 5.26% accuracy improvement, and for the VGG series, the average accuracy increased by 5.82%, outperforming previous research methods without adding additional computational costs. Further experiments show that this framework can effectively enhance the robustness and calibration of the network. • The GK-KD framework we propose enhances network performance by distilling accurate and important knowledge. • Experiments show our framework balances accuracy, computational resources, and efficiency. • Effectiveness on three datasets shown; further tests prove method's universality, reliability, and robustness. [ABSTRACT FROM AUTHOR]

To understand the mechanism of dark abiotic mercury (Hg) methylation by algae-derived dissolved organic matter (DOM) and effectively manage the environmental risks of mercury methylation in aquaculture areas, we investigated the influence of subfractions of DOM released from algae (Ulothrix sp.) decomposition on mercury methylation. The results showed that the hydrophobic basic component (HOB) in DOM exhibited the most substantial promotion effect on Hg methylation. The methylmercury (MeHg) production in the HOB treatment increased significantly, while the production rate of MeHg (%MeHg represented the concentration ratio of MeHg to THg) in the six subfractions treated solutions decreased significantly with the increase of Hg concentration. The change of the %MeHg was more evident at low Hg concentration, indicating the limited number of binding sites and methyl donors on DOM. As a consequence, Hg(Ⅱ) in the solution could not be converted into MeHg in equal proportion. Furthermore, the production of MeHg in solution was significantly reduced by the decomposed algae DOM, and its concentration was in the range of 0.017–0.085 ng·L−1 (significantly lower than undecomposed algal). The difference between the decomposed and the non-decomposed algae DOM reached a significant level (P < 0.05). When the DOM decayed for 20 and 30 days, the Hg methylation ability of DOM was weakened most obviously. During the decomposition process, considerable variations were observed among the subfractions, with HOB consistently playing a dominant role in Hg methylation. At the same time, the hydrophilic acid component exhibited a significant inhibitory effect on Hg methylation. Generally, the main components (e.g. HOB and HIA (hydrophilic acid component)) of DOM affecting mercury methylation were found in our study, which provided a better understanding of algae-derived DOM subfractions on the Hg methylation, in an attempt to prevent and control water pollution in aquaculture areas. [Display omitted] • Six DOM subfractions derived from algae samples were successfully separated with a resin-based protocol. • Methylation promoting capacity of DOM derived from algae was significantly reduced after decomposition. • Among the six subfractions sorted out, hydrophobic basic subfraction (HOB) played a dominant role in methylation promotion. • The algae rich in HOB subfraction may pose high potential to cause fish pollution by MeHg and sequent risk to human health. [ABSTRACT FROM AUTHOR]

[Display omitted] • Ni phyllosilicate was firstly synthesized by a modified microwave synthesis method. • Surfactant modification led to ultrahigh specific surface area of 535 m2 g−1 and ultrathin layer of Ni phyllosilicate of only 1.43 nm. • Hexadecyl trimethyl ammonium bromide showed higher promotion effect compared with P123 and F127. • The special structure resulted in competitive catalytic activity for CO 2 methanation. • This catalyst could exhibit high stability for 100 h in a long-term test. The traditional techniques for the synthesis of nickel phyllosilicates usually time-consuming and energy-intensive, which often lead to the formation of layers with excessive thickness due to uncontrolled crystal growth. In order to overcome these challenges, this work introduces a microwave-assisted synthesis strategy to facilitate the synthesis of Ni-phyllosilicate-based catalysts within an exceptionally short duration of only five minutes, attaining a peak temperature of merely 102 °C. To enhance the specific surface area and to increase the exposure of active sites, an investigation was conducted involving three surfactants. The employment of hexadecyl trimethyl ammonium bromide (CTAB) has yielded remarkable results, with an ultrahigh specific surface area reaching 535 m2 g−1 and an ultrathin lamellar thickness of 1.43 nm. The catalyst exhibited an impressive CO 2 conversion of 81.7 % at 400 °C, 60 L g−1 h−1, 0.1 MPa. It also demonstrated a substantial turnover frequency for CO 2 (TOF CO2) of 5.4 ± 0.1 × 10−2 s−1, alongside a relatively low activation energy (E a) of 80.74 kJ·mol−1. Moreover, the catalyst maintained its high stability over a period of 100 h and displayed high resistance to sintering. To further elucidate growth temperature gradient of the catalyst and concentration gradient of the materials involved, COMSOL Multiphysics (COMSOL) simulations were effectively utilized. In conclusion, this work breaks the limitation associated with traditional, laborious synthesis methods for Ni-phyllosilicates, which can produce materials with high surface area and thin-layer characteristics. [ABSTRACT FROM AUTHOR]

Paspalum distichum L. was tested to evaluate its ability to phytoremediate mercury (Hg) contaminated soil over a 60-d period by analysis of the total Hg concentrations in roots and leaves. Hg concentration in Hg-contamination soil decreased by 70.0 μg g−1 after 60 day of grass cultivation and Hg was readily taken up by the roots (4.51 ± 1.90 μg g−1) rather than the leaves (0.35 ± 0.02 μg g−1). In addition, a comparative proteomic study was performed to unravel the protein expression involved in the Hg stress response in P. distichum L. A total of 49 proteins were classified as differentially proteins in the roots by the 'top three' proteomic analysis, of which 32 were up-regulated and 17 down-regulated in response to Hg stress. These changed proteins were classified by gene ontology analysis into five complex molecular functions involving photosynthesis and energy metabolism (31%), oxidative stress (14%), protein folding (16%), sulfur compound metabolism (10%), metal binding, and ion transport (29%). Moreover, the protein expression patterns were consistent with the metabolism pathway results. Overall, the results contribute to our understanding of the molecular mechanisms of the Hg response in P. distichum and we propose a theoretical basis for the phytoremediation of Hg-contaminated soils. Image 1 • Forty nine abundant proteins were identified in the roots of P. disctichum. L. • Five kind proteins of P. disctichum. L were classified under the Hg stress condition. • During the plant cultivation process, mercury content in soil decreased by 31.3%. • Mercury in roots and shoots of P. distichum. L increased by 10.0 and 5.83 times. Mercury could be bioaccumulated in Paspalum distichum L and the proteomic study showed 32 proteins were up-regulated and 17 proteins were down-regulated in response to Hg stress in the roots. [ABSTRACT FROM AUTHOR]

With the ability to generate in situ real-time electric signals, electrochemically active biofilm (EAB) sensors have attracted wide attention as a promising water biotoxicity early-warning device. Organic matters serving as the electron donors potentially affect the electric signal's output and the sensitivity of the EAB sensor. To explore the influence of organic matters on EAB sensor's performance, this study tested six different organic matters during the sensor's inoculation. Besides the acetate, a conventional and widely used organic matter, propionate and lactate were also found capable of starting up the sensor. Moreover, the propionate-fed (PF) sensor delivered the highest sensitivity, which are respectively 1.4 times and 2.8 times of acetate-fed (AF) sensor and lactate-fed (LF) sensor. Further analysis revealed that EAB of PF sensor had more vulnerable intracellular metabolism than the others, which manifested as the most severe energy metabolic suppression and reactive oxygen species attack. Regarding the microbial function, a two-component system that was deemed as an environment awareness system was found in the EAB of PF, which also contributed to its high sensitivity. Finally, PF sensor was tested in real water environment to deliver early-warning signals. • Effects of organic matters on the EAB sensors' sensitivity were investigated. • Propionate rather than acetate was suit for feeding sensitive sensors. • Energy metabolic suppression and ROS attack led to high sensitivity. • Microbial function difference was closely related to the sensitivity change. • Low concentration of propionate was more suit for high sensitivity. [ABSTRACT FROM AUTHOR]

In this study, we addressed the challenge of excessive fluorescence background in bacterial colony Raman detection and aimed to achieve rapid identification of colonies. To overcome this issue, we employed a combination of droplet microcavity and label-free Surface Enhanced Raman Spectroscopy (SERS) technologies for spectroscopic analysis of five species of lactic acid bacteria (LAB) colonies during fermentation. This approach, coupled with Supported Vector Machine (SVM) and K-Nearest Neighbors (KNN) machine learning algorithms, facilitated the identification and analysis of spectral data. Comparing the results with conventional bacterial colony Raman spectra, the SERS spectra exhibited clear peaks, a higher and more stable signal-to-noise ratio, and noticeable spectral differences between various colonies, overcoming the limitations of insufficient fluorescence background. Moreover, the detection speed was notably enhanced, each SERS spectrum requires only 0.5 s, and the acquisition of the 100 spectral data points necessary for one bacterial colony is accomplished in less than 1 min. The SVM algorithm demonstrated a bacterial colony identification rate exceeding 95%, while the KNN algorithm achieved a rate surpassing 90%. These findings highlight the practical importance of using droplet microcavity combined with label-free SERS technology for quick and robust identification of the bacterial colonies. • Droplet optical microcavity combined with label-free SERS technology for direct ultra-fast identification of colonies. • The process does not require additional processing of the colonies and can complete data collection within 1 min. • Excellent results for the identification of lactic acid bacteria colonies for yogurt fermentation. [ABSTRACT FROM AUTHOR]

To investigate the effect of the organic loading rate (OLR) on anaerobic treatment of monosodium glutamate (MSG) wastewater, a lab-scale up-flow anaerobic blanket (UASB) reactor was continuously operated over a 222-day period. The overall performances of COD removal and methane recovery initially exhibited an increase and subsequently decreased when the OLR was increased from 1 g-COD/L/d to 24 g-COD/L/d. At the optimal OLR of 8 g-COD/L/d, superior performance was obtained with a maximum COD removal efficiency of 97%, a methane production rate of 2.3 L/L/d, and specific methanogenic activity of 86 mg-CH 4 /g-VSS/d (feeding on glutamate), with superior characteristics of sludge in VSS concentration, average diameter of granules, and settling velocity. According to the results of the specific methanogenic activity, the methanogenic pathway was more inclined to pass through acetate than through hydrogen. Methanosarcina instead of Methanosaeta , with Methanobacterium and greatly increased Firmicutes, dominated in the UASB reactor after long term operation. These results support that the OLR had a substantial effect on both the treatment and energy recovery efficiency of MSG wastewater as well as on microbial community variations in the UASB reactor. Image 1 • Methanogenic degradation of monosodium glutamate wastewater was validated. • Optimal COD removal and CH 4 recovery performance was obtained at OLR below 8 g/L/d. • Maximal glutamate removal capacity of the UASB reactor reached 18 g-COD/L/d. • Methanobacterium and Methanosarcina were dominated after long-term operation. • The OLR had a substantial impact on the microbial community variations in the UASB. [ABSTRACT FROM AUTHOR]

Nitrite-dependent anaerobic methane-oxidizing (n-damo) process has a promising prospect in anaerobic wastewater treatment, utilizing methane as the sole electron source to remove nitrite. However, the metabolic activity of n-damo bacteria is too low for practical application. This study aimed to stimulate n-damo process by introducing conductive nano-magnetite and/or electron shuttle anthraquinone-2,6-disulfonate (AQDS), and also set a comparative treatment of adding insulated ferrihydrite. The results showed that the nitrite reduction rate was enhanced the most significantly in treatment with nano-magnetite, approximately 1.6 times higher than that of the control without any supplement. While ferrihydrite application showed an adverse effect on n-damo process. The well-known aerobic methane oxidizer Methylomonas spp. was found to be enriched under n-damo condition with the supplementation of nano-magnetite and/or AQDS, but abundance of n-damo bacteria did not exhibit significant increase. It was hypothesized that Methylomonas spp. could be survived under anaerobic n-damo condition using oxygen produced by n-damo bacteria for the self-growth, and the nitrite reduction could be promoted through the enhancement of microbial interspecies electron transfer triggered by the introduction of conductive materials. It opens a new direction for the stimulation of n-damo activity, which needs more evidences to verify the hypothetic mechanism. • Nano-magnetite and/or AQDS stimulated the nitrite reduction of n-damo process. • Stimulation depended on acceleration of electron transfer via conductive materials. • N-damo pmoA and nod genes kept constant among different treatments. • Methylomonas was selectively enriched via nano-magnetite and/or AQDS. • A novel synergistic relationship of n-damo bacteria and Methylomonas was proposed. [ABSTRACT FROM AUTHOR]