Deep-learning models need to continually accumulate knowledge from tasks, given that the number of tasks are increasing overwhelmingly as the digital world evolves. However, standard deep-learning models are prone to ...
详细信息
Deep-learning models need to continually accumulate knowledge from tasks, given that the number of tasks are increasing overwhelmingly as the digital world evolves. However, standard deep-learning models are prone to forgetting about previously acquired skills when learning new ones. Fortunately, this catastrophic forgetting problem can be solved by means of continual learning. One popular approach in this vein is regularization-based method which penalizes parameters by giving their importance. However, a formal definition of parameter importance and theoretical analysis of regularization-based methods are elements that remain under-explored. In this paper, we first rigorously define the parameter importance by influence function, then unify the seminal methods (i.e., EWC, SI and MAS) into one whole framework. Two key theoretical results are presented in this work, and extensive experiments are conducted on standard benchmarks, which verify the superior performance of our proposed method.
Perfluorinated or polyfluorinated compounds(PFCs)continue entering to the environmental as individuals or mixtures,but their toxicological information remains largely ***,we investigated the toxic effects and ecologic...
详细信息
Perfluorinated or polyfluorinated compounds(PFCs)continue entering to the environmental as individuals or mixtures,but their toxicological information remains largely ***,we investigated the toxic effects and ecological risks of Perfluorooctane sulfonic acid(PFOS)and its substitutes on prokaryotes(Chlorella vulgaris)and eukaryotes(Microcystis aeruginosa).Based on the calculated EC50 values,the results showed that PFOS was significantly more toxic to both algae than its alternatives including Perfluorobutane sulfonic acid(PFBS)and 6:2 Fluoromodulated sulfonates(6:2 FTS),and the PFOS-PFBS mixture was more toxic to both algae than the other two PFC *** action mode of binary PFC mixtures on Chlorella vulgaris was mainly shown as antagonistic and on Microcystis aeruginosa as synergistic,by using Combination index(CI)model coupled with Monte Carlo *** mean risk quotient(RQ)value of three individual PFCs and their mixtures were all below the threshold of 10^(−1),but the risk of those binary mixtures were higher than that of PFCs individually because of their synergistic *** findings contribute to enhance the understanding of the toxicological information and ecological risks of emerging PFCs and provide a scientific basis for their pollution control.
Anode-free lithium metal batteries are prone to capacity degradation and safety hazards due to the formation and growth of lithium dendrites. The interface between the current collector and deposited lithium plays a c...
详细信息
Anode-free lithium metal batteries are prone to capacity degradation and safety hazards due to the formation and growth of lithium dendrites. The interface between the current collector and deposited lithium plays a critical role in preventing dendrite formation by regulating the thermodynamics and kinetics of lithium deposition. In this study, we develop a phase field model to investigate the influence of the current collector's surface energy on lithium deposition morphology and its effect on the quality of the lithium metal film. It is demonstrated that a higher surface energy of the current collector promotes the growth of lithium metal along the surface of the current collector. Further, our simulation results show that a higher surface energy accelerates the formation of the lithium metal film while simultaneously reducing its surface roughness. By examining different contact angles and applied potentials, we construct a phase diagram of deposition morphology, illustrating that increased surface energy facilitates the dense and uniform deposition of lithium metal by preventing the formation of lithium filaments and voids. These findings provide new insights into the development and application of anode-free lithium metal batteries.
Depth estimation is crucial in surface reconstruction for laparoscopic minimally invasive surgery (MIS) navigation. The structured light depth recovery algorithm accurately restores pixel-level depth but is time-consu...
详细信息
For mines with poor ore bodies and surrounding rocks,the general mining method does not allow the ore to be extracted from underground safely and *** these mines,the downward layered filling mining technique is undoub...
详细信息
For mines with poor ore bodies and surrounding rocks,the general mining method does not allow the ore to be extracted from underground safely and *** these mines,the downward layered filling mining technique is undoubtedly the most suitable mining *** downward filling mining technique may eliminate the troubles relating to poor ore deposit conditions,such as production safety,ore loss rate,and depletion ***,in this technique,the safety of the artificial roof of the next stratum is of paramount *** tailings backfilling(CTB)that is not sufficiently cemented and causes collapses could threaten ore *** paper explores a diamond-shaped composite structure to mimic the stability of a glued false roof in an actual infill mine based on the recently emerged three-dimensional(3D)printing *** means such as three-point bending and digital image correlation(DIC)techniques were used to explore the flexural characteristics of 3D construction specimens and CTB combinations with different cement/tailings weight ratios at diverse layer *** results show that the 3D structure with a 14-mm ply height and CTB has strong flexural characteristics,with a maximum deflection value of 30.1 mm,while the 3D-printed rhomboid polymer(3D-PRP)structure with a 26-mm ply height is slightly worse in terms of flexural strength characteristics,but it has a higher maximum flexural strength of 2.83 MPa.A combination of 3D structure and CTB has more unique mechanical properties than CTB *** research work offers practical knowledge on the artificial roof performance of the downward layered filling mining technique and builds a scientific knowledge base regarding the successful application of CTB material in mines.
Federated learning (FL) is an increasingly popular privacy-preserving collaborative machine learning paradigm that enables clients to train a global model collaboratively without sharing their raw data. Despite its ad...
详细信息
Federated learning (FL) is an increasingly popular privacy-preserving collaborative machine learning paradigm that enables clients to train a global model collaboratively without sharing their raw data. Despite its advantages, FL is vulnerable to untargeted Byzantine poisoning attacks in which malicious clients send incorrect model updates during training to disrupt the global model's performance or prevent it from converging. Existing defenses based on anomaly detection typically rely on additional auxiliary datasets and assume a known and fixed proportion of malicious clients. To overcome these shortcomings, we propose FedMP, a multi-pronged defense algorithm against untargeted Byzantine poisoning attacks. FedMP's primary idea is to detect anomalous variations in the magnitude and direction of model updates across communication rounds. In particular, FedMP first utilizes an adaptive scaling module to limit the impact of malicious updates with anomalous amplitudes. Then, FedMP identifies and filters malicious model updates with abnormal directions through dynamic clustering and partial filtering methods. Finally, FedMP extracts pure ingredients from the filtered updates as reputation scores for model aggregation to further reduce the influence of malicious updates. Comprehensive evaluations across three publicly accessible datasets demonstrate that FedMP significantly outperforms the existing Byzantine robust defenses under a high proportion of malicious clients (0.7 in our experiments) and high Non-IID degree (0.1 in our experiments) scenarios.
Bloodstains are crucial pieces of physical evidences found at violent crime scenes, providing valuable information for reconstructing forensic cases. However, there is limited data on how bloodstain lipidomes change o...
详细信息
Bloodstains are crucial pieces of physical evidences found at violent crime scenes, providing valuable information for reconstructing forensic cases. However, there is limited data on how bloodstain lipidomes change over time after deposition. Hence, we deployed a high-throughput high-performance liquid chromatography-mass spectrometry (HPLC-MS) approach to construct lipidomic atlases of bloodstains, whole blood, plasma, and blood cells from 15 healthy adults. A time-course analysis was also performed on bloodstains deposited for up to 6 months at room temperature (similar to 25 degrees C). The molecular levels of 60 out of 400 detected lipid species differed dramatically between bloodstain and whole blood samples, with major disturbances observed in membrane glycerophospholipids. More than half of these lipids were prevalent in the cellular and plasmic fractions;approximately 27% and 10% of the identified lipids were uniquely derived from blood cells and plasma, respectively. Furthermore, a subset of 65 temporally dynamic lipid species arose across the 6-month room-temperature deposition period, with decreased triacylglycerols (TAGs) and increased lysophosphatidylcholines (LPCs) as representatives, accounting for approximately 8% of the total investigated lipids. The instability of lipids increased linearly with time, with the most variability observed in the first 10 days. This study sheds light on the impact of air-drying bloodstains on blood components at room temperature and provides a list of potential bloodstain lipid markers for determining the age of bloodstains.
Database watermarking technologies provide an effective solution to data security problems by embedding the watermark in the database to prove copyright or trace the source of data ***,when the watermarked database is...
详细信息
Database watermarking technologies provide an effective solution to data security problems by embedding the watermark in the database to prove copyright or trace the source of data ***,when the watermarked database is used for data mining model building,such as decision trees,it may cause a different mining result in comparison with the result from the original database caused by the distortion of watermark *** watermarking algorithms mainly consider the statistical distortion of data,such as the mean square error,but very few consider the effect of the watermark on database ***,in this paper,a consistency preserving database watermarking algorithm is proposed for decision ***,label classification statistics and label state transfer methods are proposed to adjust the watermarked data so that the model structure of the watermarked decision tree is the same as that of the original decision ***,the splitting values of the decision tree are adjusted according to the defined constraint ***,the adjusted database can obtain a decision tree consistent with the original decision *** experimental results demonstrated that the proposed algorithm does not corrupt the watermarks,and makes the watermarked decision tree consistent with the original decision tree with a small distortion.
This study systematically analyzes the influence of different combined joint dip angles on rock mass failure modes and damage mechanisms through uniaxial compression tests on granite specimens with prefabricated Y-sha...
详细信息
This study systematically analyzes the influence of different combined joint dip angles on rock mass failure modes and damage mechanisms through uniaxial compression tests on granite specimens with prefabricated Y-shaped discontinuities,combined with digital speckle and acoustic emission(AE)*** results show that as the dip angle of the primary joint increases,the failure mode transitions from overall failure to wedge block ejection and shear failure.A failure mode identification model was established based on main crack dip angle thresholds(40°,45°),uniaxial compressive strength thresholds(40,90 MPa),and energy core zone proportion thresholds(20%,10%),achieving an accuracy of 93.3%.In the overall failure and wedge block ejection modes,a sharp increase in shear crack ratio and a sudden drop in the acoustic emission b-value occur in the high-stress phase(>0.6σ_(c)),while in the shear failure mode,significant fluctuations are observed due to the shear-tension alternation,making it difficult to identify a single critical ***,joint slip in the overall failure and wedge block ejection modes primarily occurs during the failure instability phase(>0.8σ_(c)).These findings provide theoretical support for stability evaluation of complex fractured rock masses and practical guidance for engineering safety construction.
To investigate the microstructure and creep properties of a hot corrosion resistant Ni-based single crystal superal oy containing different hafnium(Hf)additions(0-0.4wt.%),creep test was performed at 980℃/200 *** mic...
详细信息
To investigate the microstructure and creep properties of a hot corrosion resistant Ni-based single crystal superal oy containing different hafnium(Hf)additions(0-0.4wt.%),creep test was performed at 980℃/200 *** microscopy,scanning electron microscopy,electron probe micro analysis(EPMA),and transmission electron microscopy were employed to analyze the microstructure *** the increase of Hf,the creep rupture life of the alloys at 980℃/200 MPa gradually *** analysis reveals that Hf promotes the transformation of carbide morphology from script to rod-like and finally to *** the addition of Hf,there is an increase in the volume fraction of blocky MC carbides,along with an elevation in the partitioning ratio of Cr and Mo ***,theγ/γ′interfacial dislocation spacing undergoes a *** is found that script carbides are more likely to cause stress concentration in high temperature creep,leading to nucleation and propagation of *** formation mechanism of blocky MC carbides is related to the increase in precipitation temperature and lattice constant,and its beneficial impact on creep resistance is also investigated based on the analysis of the creep test results.
暂无评论