Magnesium(Mg)alloys are widely used lightweight structural materials for automobiles and help reduce carbon emissions.However,their use increases the production of Mg alloy scrap,which is recycled at a much lower rate...Magnesium(Mg)alloys are widely used lightweight structural materials for automobiles and help reduce carbon emissions.However,their use increases the production of Mg alloy scrap,which is recycled at a much lower rate than aluminum,and its greater complexity poses challenges to existing recycling processes.Although vacuum distillation can be used to recycle Mg alloy scrap,this requires optimizing and maximizing metal recirculation,but there has been no thermodynamic analysis of this process.In this study,the feasibility and controllability of separating inclusions and 23 metal impurities were evaluated,and their distribution and removal limits were quantified.Thermodynamic analyses and experimental results showed that inclusions and impurity metals of separation coefficient lgβ_(i)≤-5,including Cu,Fe,Co,and Ni below 0.001 ppm,could be removed from the matrix.All Zn entered the recycled Mg,while impurities with-1<lgβ_(i)<-5 such as Li,Ca,and Mn severely affected the purity of the recycled Mg during the later stage of distillation.Therefore,an optimization strategy for vacuum distillation recycling:lower temperatures and higher system pressures for Zn separation in the early stage,and the early termination of the recovery process in the later stage or a continuous supply of raw melt can also prevent contamination during recycling.The alloying elements Al and Zn in Mg alloy scrap can be further recovered and purified by vacuum distillation when economically feasible,to maximize the recycling of metal resources.展开更多
High-purity indium finds extensive application in the aerospace,electronics,medical,energy,and national defense sectors.Its purity and impurity contents significantly influence its performance in these applications.Hi...High-purity indium finds extensive application in the aerospace,electronics,medical,energy,and national defense sectors.Its purity and impurity contents significantly influence its performance in these applications.High-purity indium was prepared by combining zone refining with vacuum distillation.Results show that the average removal efficiency of impurity Sb can approach 95%,while the removal efficiency of impurities Sn and Bi can reach over 95%,and the removal efficiency of Si,Fe,Ni,and Pb can reach over 85%.Ultimately,the amount of Sn and Sb impurities is reduced to 2.0 and 4.1μg/kg,respectively,and that of most impurities,including Fe,Ni,Pb,and Bi,is reduced to levels below the instrumental detection limit.The average impurity removal efficiency is 90.9%,and the indium purity reaches 7N9.展开更多
Nervonic acid(NA) is a long-chain monounsaturated fatty acid with significant potential for neural fiber repair.In this study,a mixed fatty acid methyl ester was synthesized as the raw material through saponification ...Nervonic acid(NA) is a long-chain monounsaturated fatty acid with significant potential for neural fiber repair.In this study,a mixed fatty acid methyl ester was synthesized as the raw material through saponification of Acer truncatum Bunge seed oil.Based on the differences in boiling points and relative volatilities of various components,a four-stage vacuum batch distillation process was employed to enrich the nervonic acid methyl ester(NAME).The effect of distillation process parameters on enrichment efficiency was investigated,including distillation temperature,operating pressure,and reflux ratio.The purity of NAME achieved as 91.20% under optimal conditions and the corresponding yield was 48.91%.To further increase the purity,a low-temperature crystallization process was adopted and a final purity of NAME was obtained as 97.56%.Simulation of the above four-stage batch distillation was conducted using Aspen Plus software,and a continuous distillation processes was further simulated to establish a theoretical basis for future industrial-scale production.The results of experiments and simulation demonstrate that the integrated process of vacuum distillation and low-temperature crystallization exhibits remarkable separation performances,providing robust guidance for the production of high-purity NA.展开更多
Hydrophobic nanofiber composite membranes comprising polyimide and metal-organic frameworks are developed for desalination via direct contact membrane distillation(DCMD).Our study demonstrates the synthesis of hydroph...Hydrophobic nanofiber composite membranes comprising polyimide and metal-organic frameworks are developed for desalination via direct contact membrane distillation(DCMD).Our study demonstrates the synthesis of hydrophobic polyimides with trifluoromethyl groups,along with superhydrophobic UiO-66(hMOF)prepared by phenylsilane modification on the metal-oxo nodes.These components are then combined to create nanofiber membranes with improved hydro ph obi city,ensuring long-term stability while preserving a high water flux.Integration of hMOF into the polymer matrix further increases membrane hydrophobic properties and provides additional pathways for vapor transport during MD.The resulting nanofiber composite membranes containing 20 wt%of hMOFs(PI-1-hMOF-20)were able to desalinate hypersaline feed solution of up to 17 wt%NaCl solution,conditions that are beyond the capability of reverse osmosis systems.These membranes demonstrated a water flux of 68.1 kg m^(-2)h^(-1) with a rejection rate of 99.98%for a simulated seawater solution of 3.5 wt%NaCl at 70℃,while maintaining consistent desalination performance for 250 h.展开更多
Traditional biodiesel production primarily uses methanol as the acyl acceptor,but its toxicity to lipase increases process complexity and operational difficulty elevate manufacturing costs.This study aimed to explore ...Traditional biodiesel production primarily uses methanol as the acyl acceptor,but its toxicity to lipase increases process complexity and operational difficulty elevate manufacturing costs.This study aimed to explore a new method for enzymatic synthesis of biodiesel with methyl methacrylate(MMA)as acyl acceptor.Meanwhile,a 1,3-position specific lipase Lipozyme RM IM was applied as biocatalyst,which enables simultaneous production of biodiesel(FAMEs)and methacrylate fatty acid glycerides(MFAGs)via specific sn-1,3 transesterification of MMA with triglyceride.Under the optimal reaction conditions:temperature of 50℃,molar ratio of 4:1 for MMA to triglyceride,enzyme dosage of 7.5%(mass),and an extra water addition of 0.5%(mass);triglyceride conversion rate of 97%,and FAMEs yield of 65%could be obtained.Simultaneously,the multistage short-path distillation and column chromatographic method were combined used for the separation of the mixed products.Finally,the purity of FAME,MFADG,DMFAG,and MMFAG were 98%,97.8%,95.3%,and 81.78%,respectively.In this new approach,MMA demonstrates lower toxicity to lipases,allowing for straightfo rward addition of all the substrates without complex addition process,and enhancing operational feasibility.Meanwhile,the by-products of MFAGs could be applied as monomers in varnishes and protective coatings,which increased the value of the products.Thus,this investigation providing an alternative way to produce biodiesel,and providing a new pathway for the sustainable development of biodiesel.展开更多
Acetone-butanol-ethanol(ABE)fermentation is a primary strategy for producing bio-based n-butanol from abundant renewable biomass.In the typical ABE production chain,distillation is an essential unit for high purity AB...Acetone-butanol-ethanol(ABE)fermentation is a primary strategy for producing bio-based n-butanol from abundant renewable biomass.In the typical ABE production chain,distillation is an essential unit for high purity ABE productions,but has long been criticized by the energy-inefficient processes due to the extremely low solvents concentration received in the upstream fermentation system.Over the past decades,efforts have been dedicated to developing eco-efficient ABE distillation processes aimed at reducing both energy costs and capital investments.In this review,a comprehensive overview on ABE distillation systems is provided from physico-chemical properties in feed and thermodynamics to the process constructions and applications.The recent trends in distillation sequence construction that fitting with the rapid developed upstream in situ product recovery(ISPR)systems are emphasized.Furthermore,towards developing a more efficient ABE distillation system,the review takes a broad overview of the intensification strategies for ABE distillation.Along with systematic introduction of the key examples,the future directions for ABE distillation techniques development are also discussed towards a sustainable and low-carbon emission biorefineries.展开更多
Membrane distillation(MD)has gained extensive attention for treating highly saline wastewater.However,membrane scaling during the MD process has hindered the rapid development of this technology.Current approaches to ...Membrane distillation(MD)has gained extensive attention for treating highly saline wastewater.However,membrane scaling during the MD process has hindered the rapid development of this technology.Current approaches to mitigate scaling in membrane distillation focus primarily on achieving enhanced hydrophobicity and even superhydrophobicity via utilizing fluorinated fibrous membrane or introducing perfluorosilane modification.Considering the environmental hazards posed by fluorinated compounds,it is highly desirable to develop non-fluorinated membranes with enhanced anti-scaling properties for effective membrane distillation.In this study,we present a non-fluorinated liquid-like MD membrane with exceptional anti-scaling performance.This membrane was facilely fabricated by grafting linear polydimethylsiloxane(LPDMS)onto a hydrophilic polyether sulfone(PES)membrane pre-coated with the intermediate layers of polydopamine and silica(denoted as LPDMS-PES).Remarkably,LPDMS-PES manifested a drastically improved scaling resistance in continuous MD tests than its perfluorinated counterpart,i.e.,1H,1H,2H,2H-perfluorooctyltrichlorosilane-modified PES membrane(PFOS-PES),in both heterogeneous nucleation-dominated and crystal deposition-dominated scaling processes,despite the latter having a smaller surface energy.LPDMS-PES demonstrated a reduction of crystal accumulation of approximately 85%for Na Cl and 73%for Ca SO_(4) in the heterogeneous nucleation-dominated scaling process compared to PFOS-PES.Additionally,in the crystal deposition-dominated scaling process LPDMS-PES exhibited a reduction of about 70%in scale accumulation.These results explicitly evidenced the great potential of the liquid-like membrane to minimize scaling in membrane distillation by inhibiting both scale nucleation and adhesion onto the membrane.We believe the findings of this study have important implications for the design of high-performance MD membranes,particularly in the quest for environmentally sustainable alternatives to perfluorinated materials.展开更多
The operational state of distillation columns significantly impacts product quality and production efficiency.However,due to the complex operation and diverse influencing factors,ensuring the safety and efficient oper...The operational state of distillation columns significantly impacts product quality and production efficiency.However,due to the complex operation and diverse influencing factors,ensuring the safety and efficient operation of the distillation columns becomes paramount.This research combines passive acoustic monitoring with artificial intelligence techniques,proposed a technology based on residual network(ResNet),which involves the transformation of the acoustic signals emitted by three distillation columns under different operating states.The acoustic signals were initially in one-dimensional waveform format and then converted into two-dimensional Mel-Frequency Cepstral Coefficients spectrogram database using fast Fourier transform.Ultimately,this database was employed to train a ResNet for the purpose of identifying the operational states of the distillation columns.Through this approach,the operational states of distillation columns were monitored.Various faults,including flooding,entrainment,dry-tray,etc.,were diagnosed with an accuracy of 98.91%.Moreover,an intermediate transitional state between normal operation and fault was identified and accurately recognized by the proposed method.Under the transitional state,the acoustic signals achieved an accuracy of 97.85%on the ResNet,which enables early warnings before faults occur,enhancing the safety of chemical production processes.The approach presents a powerful tool for the monitoring and diagnosis of chemical equipment,particularly distillation columns,ensuring the safety and efficiency.展开更多
To address the challenge of recognizing small target information on traffic panels,a model named MLGIA is proposed based on PaddlePaddle.MLGIA is composed of MobilenetV3 with lightweight GhostBlock(LGB)and an improved...To address the challenge of recognizing small target information on traffic panels,a model named MLGIA is proposed based on PaddlePaddle.MLGIA is composed of MobilenetV3 with lightweight GhostBlock(LGB)and an improved augmented feature pyramid network(IAFPN).In this model,LGB improves MobilenetV3 by optimizing the convolutional structure and employing linear transformations to extract sufficient feature maps;IAFPN enhances feature representation through pruning techniques and channel-reduction convolutions.Additionally,knowledge distillation compresses the model and improves its accuracy,while the match category information(MCI)method further optimizes the processing of the detected category information.Experimental results demonstrate that MLGIA outperforms MobilenetV3.MLGIA achieves a detection accuracy comparable to YOLOv8n,with significantly lower resource consumption.Therefore,MLGIA is a strong complement in the traffic panel information recognition domain.展开更多
The production of high-purity propylene glycol monomethyl ether acetate(PMA)through the transesterification of propylene glycol monomethyl ether(PM)and methyl acetate(MeOAc)is traditionally catalyzed by sodium methoxi...The production of high-purity propylene glycol monomethyl ether acetate(PMA)through the transesterification of propylene glycol monomethyl ether(PM)and methyl acetate(MeOAc)is traditionally catalyzed by sodium methoxide.However,the practical application of this method is significantly hindered by the inherent limitations of sodium methoxide,such as its high sensitivity to moisture and propensity for solid precipitation,which impede its effective use in continuous processes.This work proposed a continuous catalytic distillation(CD)process utilizing Amberlyst 15 cation exchange resin as the catalyst.A comprehensive series of reaction kinetic and CD experiments were conducted to evaluate the performance of the proposed process.The results demonstrate that under the optimal operating conditions,namely an ester-to-ether molar ratio of 6:1,a refluxratio of 5:1,a total feed rate of 0.92 g‧min^(-1),and an evaporation rate of 266.47 m^(3)‧m^(-2)‧h^(-1),the conversion rate of PM achieves 99.95%,and the PMA yield is 97.31%.Based on these findings,a process flowsheet for a continuous CD process tailored for the production of electronic-grade PMA is presented.This design incorporates light and heavy removal steps to ensure the production of PMA with a purity of 99.99%.Additionally,the process utilizes pressure swing distillation to recover MeOAc,thereby enhancing the overall efficiencyand sustainability of the production process.The proposed continuous CD process offers a highly efficient,cost-effective,and environmentally sustainable solution for the production of electronic-grade PMA.展开更多
Optical and hybrid convolutional neural networks(CNNs)recently have become of increasing interest to achieve low-latency,low-power image classification,and computer-vision tasks.However,implementing optical nonlineari...Optical and hybrid convolutional neural networks(CNNs)recently have become of increasing interest to achieve low-latency,low-power image classification,and computer-vision tasks.However,implementing optical nonlinearity is challenging,and omitting the nonlinear layers in a standard CNN comes with a significant reduction in accuracy.We use knowledge distillation to compress modified AlexNet to a single linear convolutional layer and an electronic backend(two fully connected layers).We obtain comparable performance with a purely electronic CNN with five convolutional layers and three fully connected layers.We implement the convolution optically via engineering the point spread function of an inverse-designed meta-optic.Using this hybrid approach,we estimate a reduction in multiply-accumulate operations from 17M in a conventional electronic modified AlexNet to only 86 K in the hybrid compressed network enabled by the optical front end.This constitutes over 2 orders of magnitude of reduction in latency and power consumption.Furthermore,we experimentally demonstrate that the classification accuracy of the system exceeds 93%on the MNIST dataset of handwritten digits.展开更多
Accurate prediction of perovskite photovoltaic materials’optoelectronic properties is crucial for developing efficient and stable materials,advancing solar technology.To address poor interpretability,high computation...Accurate prediction of perovskite photovoltaic materials’optoelectronic properties is crucial for developing efficient and stable materials,advancing solar technology.To address poor interpretability,high computational complexity,and inaccurate predictions in relevant machine learningmodels,this paper proposes a novelmethodology.The technical route of this papermainly centers on the randomforest-knowledge distillation-bidirectional gated recurrent unit with attention technology(namely RF-KD-BIGRUA),which is applied in perovskite photovoltaic materials.Primarily,it combines random forest to quantitatively assess feature importance,selecting variables with significant impacts on photoelectric conversion efficiency.Subsequently,statistical techniques analyze the weight distribution of variables influencing power conversion efficiency(PCE,%)to extract key features.In the model optimization phase,knowledge distillation transfers features from complex teacher models to student models,enhancing prediction accuracy.Additionally,Bidirectional Gated Recurrent Unit with Attention technology(BiGRU-Attention)is introduced to further optimize predictive performancewhile substantially reducing computational costs.The results demonstrate that integrating statistical techniques into intelligent optimization models can quantify photovoltaic system uncertainties and reduce prediction errors before experimental fabrication,enabling efficient pre-fabrication screening of perovskite materials that meet energy-storage criteria and providing accurate guidance for material selection.展开更多
The rapid growth of digital data necessitates advanced natural language processing(NLP)models like BERT(Bidi-rectional Encoder Representations from Transformers),known for its superior performance in text classificati...The rapid growth of digital data necessitates advanced natural language processing(NLP)models like BERT(Bidi-rectional Encoder Representations from Transformers),known for its superior performance in text classification.However,BERT’s size and computational demands limit its practicality,especially in resource-constrained settings.This research compresses the BERT base model for Bengali emotion classification through knowledge distillation(KD),pruning,and quantization techniques.Despite Bengali being the sixth most spoken language globally,NLP research in this area is limited.Our approach addresses this gap by creating an efficient BERT-based model for Bengali text.We have explored 20 combinations for KD,quantization,and pruning,resulting in improved speedup,fewer parameters,and reduced memory size.Our best results demonstrate significant improvements in both speed and efficiency.For instance,in the case of mBERT,we achieved a 3.87×speedup and 4×compression ratio with a combination of Distil+Prune+Quant that reduced parameters from 178 to 46 M,while the memory size decreased from 711 to 178 MB.These results offer scalable solutions for NLP tasks in various languages and advance the field of model compression,making these models suitable for real-world applications in resource-limited environments.展开更多
Computer-vision and deep-learning techniques are widely applied to detect,monitor,and assess pavement conditions including road crack detection.Traditional methods fail to achieve satisfactory accuracy and generalizat...Computer-vision and deep-learning techniques are widely applied to detect,monitor,and assess pavement conditions including road crack detection.Traditional methods fail to achieve satisfactory accuracy and generalization performance in for crack detection.Complex network model can generate redundant feature maps and computational complexity.Therefore,this paper proposes a novel model compression framework based on deep learning to detect road cracks,which can improve the detection efficiency and accuracy.A distillation loss function is proposed to compress the teacher model,followed by channel pruning.Meanwhile,a multi-dilation model is proposed to improve the accuracy of the model pruned.The proposed method is tested on the public database CrackForest dataset(CFD).The experimental results show that the proposed method is more efficient and accurate than other state-of-art methods.展开更多
Graph neural networks(GNNs)have demonstrated excellent performance in graph representation learning.However,as the volume of graph data grows,issues related to cost and efficiency become increasingly prominent.Graph d...Graph neural networks(GNNs)have demonstrated excellent performance in graph representation learning.However,as the volume of graph data grows,issues related to cost and efficiency become increasingly prominent.Graph distillation methods address this challenge by extracting a smaller,reduced graph,ensuring that GNNs trained on both the original and reduced graphs show similar performance.Existing methods,however,primarily optimize the feature matrix of the reduced graph and rely on correlation information from GNNs,while neglecting the original graph’s structure and redundant nodes.This often results in a loss of critical information within the reduced graph.To overcome this limitation,we propose a graph distillation method guided by network symmetry.Specifically,we identify symmetric nodes with equivalent neighborhood structures and merge them into“super nodes”,thereby simplifying the network structure,reducing redundant parameter optimization and enhancing training efficiency.At the same time,instead of relying on the original node features,we employ gradient descent to match optimal features that align with the original features,thus improving downstream task performance.Theoretically,our method guarantees that the reduced graph retains the key information present in the original graph.Extensive experiments demonstrate that our approach achieves significant improvements in graph distillation,exhibiting strong generalization capability and outperforming existing graph reduction methods.展开更多
Defect detection based on computer vision is a critical component in ensuring the quality of industrial products.However,existing detection methods encounter several challenges in practical applications,including the ...Defect detection based on computer vision is a critical component in ensuring the quality of industrial products.However,existing detection methods encounter several challenges in practical applications,including the scarcity of labeled samples,limited adaptability of pre-trained models,and the data heterogeneity in distributed environments.To address these issues,this research proposes an unsupervised defect detection method,FLAME(Federated Learning with Adaptive Multi-Model Embeddings).The method comprises three stages:(1)Feature learning stage:this work proposes FADE(Feature-Adaptive Domain-Specific Embeddings),a framework employs Gaussian noise injection to simulate defective patterns and implements a feature discriminator for defect detection,thereby enhancing the pre-trained model’s industrial imagery representation capabilities.(2)Knowledge distillation co-training stage:a multi-model feature knowledge distillation mechanism is introduced.Through feature-level knowledge transfer between the global model and historical local models,the current local model is guided to learn better feature representations from the global model.The approach prevents local models from converging to local optima and mitigates performance degradation caused by data heterogeneity.(3)Model parameter aggregation stage:participating clients utilize weighted averaging aggregation to synthesize an updated global model,facilitating efficient knowledge consolidation.Experimental results demonstrate that FADE improves the average image-level Area under the Receiver Operating Characteristic Curve(AUROC)by 7.34%compared to methods directly utilizing pre-trained models.In federated learning environments,FLAME’s multi-model feature knowledge distillation mechanism outperforms the classic FedAvg algorithm by 2.34%in average image-level AUROC,while exhibiting superior convergence properties.展开更多
The Heterogeneous Capacitated Vehicle Routing Problem(HCVRP),which involves efficiently routing vehicles with diverse capacities to fulfill various customer demands at minimal cost,poses an NP-hard challenge in combin...The Heterogeneous Capacitated Vehicle Routing Problem(HCVRP),which involves efficiently routing vehicles with diverse capacities to fulfill various customer demands at minimal cost,poses an NP-hard challenge in combinatorial optimization.Recently,reinforcement learning approaches such as 2D Array Pointer Networks(2D-Ptr)have demonstrated remarkable speed in decision-making by modeling multiple agents’concurrent choices as a sequence of consecutive actions.However,these learning-based models often struggle with generalization,meaning they cannot seamlessly adapt to new scenarios with varying numbers of vehicles or customers without retraining.Inspired by the potential of multi-teacher knowledge distillation to harness diverse knowledge from multiple sources and craft a comprehensive student model,we propose to enhance the generalization capability of 2D-Ptr through Multiple Teacher-forcing Knowledge Distillation(MTKD).We initially train 12 unique 2D-Ptr models under various settings to serve as teacher models.Subsequently,we randomly sample a teacher model and a batch of problem instances,focusing on those where the chosen teacher performed best.This teacher model then solves these instances,generating high-reward action sequences to guide knowledge transfer to the student model.We conduct rigorous evaluations across four distinct datasets,each comprising four HCVRP instances of varying scales.Our empirical findings underscore the proposed method superiority over existing learning-based methods in terms of both computational efficiency and solution quality.展开更多
Knowledge distillation(KD)is an emerging model compression technique for learning compact object detector models.Previous KD often focused solely on distilling from the logits layer or the feature intermediate layers,...Knowledge distillation(KD)is an emerging model compression technique for learning compact object detector models.Previous KD often focused solely on distilling from the logits layer or the feature intermediate layers,which may limit the comprehensive learning of the student network.Additionally,the imbalance between the foreground and background also affects the performance of the model.To address these issues,this paper employs feature-based distillation to enhance the detection performance of the bounding box localization part,and logit-based distillation to improve the detection performance of the category prediction part.Specifically,for the intermediate layer feature distillation,we introduce feature resampling to reduce the risk of the student model merely imitating the teacher model.At the same time,we incorporate a Spatial Attention Mechanism(SAM)to highlight the foreground features learned by the student model.In terms of output layer feature distillation,we divide the traditional distillation targets into target-class objects and non-target-class objects,aiming to improve overall distillation performance.Furthermore,we introduce a one-to-many matching distillation strategy based on Feature Alignment Module(FAM),which further enhances the studentmodel’s feature representation ability,making its feature distribution closer to that of the teacher model,and thus demonstrating superior localization and classification capabilities in object detection tasks.Experimental results demonstrate that our proposedmethodology outperforms conventional distillation techniques in terms of object detecting performance.展开更多
In recent years,deep learning has made significant advancements in skin cancer diagnosis.However,most methods prioritize high prediction accuracy without considering the limitations of computational resources,making t...In recent years,deep learning has made significant advancements in skin cancer diagnosis.However,most methods prioritize high prediction accuracy without considering the limitations of computational resources,making them impractical for wearable devices.In this case,knowledge distillation has emerged as an effective method,capable of significantly reducing a model’s reliance on computational and storage resources.Nonetheless,previous research suffers from two limitations:1)the student model can only passively receive knowledge from the teacher model,and 2)the teacher model does not effectively model sample relationships during training,potentially hindering the effective transfer of sample relationship-related knowledge during knowledge distillation.To address these issues,we employ two identical student models,each equipped with a sample relationship module.This design ensures that the student models can mutually learn while modeling sample relationships.We conducted extensive experiments on the ISIC 2019 dataset to validate the effectiveness of our method.The results demonstrate that our approach significantly improves the recognition of various types of skin diseases.Compared to state-of-the-art methods,our approach exhibits higher accuracy and better generalization capabilities.展开更多
In order to improve the reconstruction accuracy of magnetic resonance imaging(MRI),an accurate natural image compressed sensing(CS)reconstruction network is proposed,which combines the advantages of model-based and de...In order to improve the reconstruction accuracy of magnetic resonance imaging(MRI),an accurate natural image compressed sensing(CS)reconstruction network is proposed,which combines the advantages of model-based and deep learning-based CS-MRI methods.In theory,enhancing geometric texture details in linear reconstruction is possible.First,the optimization problem is decomposed into two problems:linear approximation and geometric compensation.Aimed at the problem of image linear approximation,the data consistency module is used to deal with it.Since the processing process will lose texture details,a neural network layer that explicitly combines image and frequency feature representation is proposed,which is named butterfly dilated geometric distillation network.The network introduces the idea of butterfly operation,skillfully integrates the features of image domain and frequency domain,and avoids the loss of texture details when extracting features in a single domain.Finally,a channel feature fusion module is designed by combining channel attention mechanism and dilated convolution.The attention of the channel makes the final output feature map focus on the more important part,thus improving the feature representation ability.The dilated convolution enlarges the receptive field,thereby obtaining more dense image feature data.The experimental results show that the peak signal-to-noise ratio of the network is 5.43 dB,5.24 dB and 3.89 dB higher than that of ISTA-Net+,FISTA and DGDN networks on the brain data set with a Cartesian sampling mask CS ratio of 10%.展开更多
文摘Magnesium(Mg)alloys are widely used lightweight structural materials for automobiles and help reduce carbon emissions.However,their use increases the production of Mg alloy scrap,which is recycled at a much lower rate than aluminum,and its greater complexity poses challenges to existing recycling processes.Although vacuum distillation can be used to recycle Mg alloy scrap,this requires optimizing and maximizing metal recirculation,but there has been no thermodynamic analysis of this process.In this study,the feasibility and controllability of separating inclusions and 23 metal impurities were evaluated,and their distribution and removal limits were quantified.Thermodynamic analyses and experimental results showed that inclusions and impurity metals of separation coefficient lgβ_(i)≤-5,including Cu,Fe,Co,and Ni below 0.001 ppm,could be removed from the matrix.All Zn entered the recycled Mg,while impurities with-1<lgβ_(i)<-5 such as Li,Ca,and Mn severely affected the purity of the recycled Mg during the later stage of distillation.Therefore,an optimization strategy for vacuum distillation recycling:lower temperatures and higher system pressures for Zn separation in the early stage,and the early termination of the recovery process in the later stage or a continuous supply of raw melt can also prevent contamination during recycling.The alloying elements Al and Zn in Mg alloy scrap can be further recovered and purified by vacuum distillation when economically feasible,to maximize the recycling of metal resources.
基金National Key Research and Development Program of China(2023YFC2907904)National Natural Science Foundation of China(52374364)。
文摘High-purity indium finds extensive application in the aerospace,electronics,medical,energy,and national defense sectors.Its purity and impurity contents significantly influence its performance in these applications.High-purity indium was prepared by combining zone refining with vacuum distillation.Results show that the average removal efficiency of impurity Sb can approach 95%,while the removal efficiency of impurities Sn and Bi can reach over 95%,and the removal efficiency of Si,Fe,Ni,and Pb can reach over 85%.Ultimately,the amount of Sn and Sb impurities is reduced to 2.0 and 4.1μg/kg,respectively,and that of most impurities,including Fe,Ni,Pb,and Bi,is reduced to levels below the instrumental detection limit.The average impurity removal efficiency is 90.9%,and the indium purity reaches 7N9.
基金supported by the National Natural Science Foundation of China(22125802,22108150,22338001)。
文摘Nervonic acid(NA) is a long-chain monounsaturated fatty acid with significant potential for neural fiber repair.In this study,a mixed fatty acid methyl ester was synthesized as the raw material through saponification of Acer truncatum Bunge seed oil.Based on the differences in boiling points and relative volatilities of various components,a four-stage vacuum batch distillation process was employed to enrich the nervonic acid methyl ester(NAME).The effect of distillation process parameters on enrichment efficiency was investigated,including distillation temperature,operating pressure,and reflux ratio.The purity of NAME achieved as 91.20% under optimal conditions and the corresponding yield was 48.91%.To further increase the purity,a low-temperature crystallization process was adopted and a final purity of NAME was obtained as 97.56%.Simulation of the above four-stage batch distillation was conducted using Aspen Plus software,and a continuous distillation processes was further simulated to establish a theoretical basis for future industrial-scale production.The results of experiments and simulation demonstrate that the integrated process of vacuum distillation and low-temperature crystallization exhibits remarkable separation performances,providing robust guidance for the production of high-purity NA.
基金supported by the Australian Research Council Discovery Early Career Researcher Award Scheme(DE220100135 and DE220100435)。
文摘Hydrophobic nanofiber composite membranes comprising polyimide and metal-organic frameworks are developed for desalination via direct contact membrane distillation(DCMD).Our study demonstrates the synthesis of hydrophobic polyimides with trifluoromethyl groups,along with superhydrophobic UiO-66(hMOF)prepared by phenylsilane modification on the metal-oxo nodes.These components are then combined to create nanofiber membranes with improved hydro ph obi city,ensuring long-term stability while preserving a high water flux.Integration of hMOF into the polymer matrix further increases membrane hydrophobic properties and provides additional pathways for vapor transport during MD.The resulting nanofiber composite membranes containing 20 wt%of hMOFs(PI-1-hMOF-20)were able to desalinate hypersaline feed solution of up to 17 wt%NaCl solution,conditions that are beyond the capability of reverse osmosis systems.These membranes demonstrated a water flux of 68.1 kg m^(-2)h^(-1) with a rejection rate of 99.98%for a simulated seawater solution of 3.5 wt%NaCl at 70℃,while maintaining consistent desalination performance for 250 h.
文摘Traditional biodiesel production primarily uses methanol as the acyl acceptor,but its toxicity to lipase increases process complexity and operational difficulty elevate manufacturing costs.This study aimed to explore a new method for enzymatic synthesis of biodiesel with methyl methacrylate(MMA)as acyl acceptor.Meanwhile,a 1,3-position specific lipase Lipozyme RM IM was applied as biocatalyst,which enables simultaneous production of biodiesel(FAMEs)and methacrylate fatty acid glycerides(MFAGs)via specific sn-1,3 transesterification of MMA with triglyceride.Under the optimal reaction conditions:temperature of 50℃,molar ratio of 4:1 for MMA to triglyceride,enzyme dosage of 7.5%(mass),and an extra water addition of 0.5%(mass);triglyceride conversion rate of 97%,and FAMEs yield of 65%could be obtained.Simultaneously,the multistage short-path distillation and column chromatographic method were combined used for the separation of the mixed products.Finally,the purity of FAME,MFADG,DMFAG,and MMFAG were 98%,97.8%,95.3%,and 81.78%,respectively.In this new approach,MMA demonstrates lower toxicity to lipases,allowing for straightfo rward addition of all the substrates without complex addition process,and enhancing operational feasibility.Meanwhile,the by-products of MFAGs could be applied as monomers in varnishes and protective coatings,which increased the value of the products.Thus,this investigation providing an alternative way to produce biodiesel,and providing a new pathway for the sustainable development of biodiesel.
基金funded by the National Natural Science Foundation of China(22078018)the Natural Science Foundation of Beijing(2222016).
文摘Acetone-butanol-ethanol(ABE)fermentation is a primary strategy for producing bio-based n-butanol from abundant renewable biomass.In the typical ABE production chain,distillation is an essential unit for high purity ABE productions,but has long been criticized by the energy-inefficient processes due to the extremely low solvents concentration received in the upstream fermentation system.Over the past decades,efforts have been dedicated to developing eco-efficient ABE distillation processes aimed at reducing both energy costs and capital investments.In this review,a comprehensive overview on ABE distillation systems is provided from physico-chemical properties in feed and thermodynamics to the process constructions and applications.The recent trends in distillation sequence construction that fitting with the rapid developed upstream in situ product recovery(ISPR)systems are emphasized.Furthermore,towards developing a more efficient ABE distillation system,the review takes a broad overview of the intensification strategies for ABE distillation.Along with systematic introduction of the key examples,the future directions for ABE distillation techniques development are also discussed towards a sustainable and low-carbon emission biorefineries.
基金supported by National Natural Science Foundation of China(Nos.22072185,12072381)Guangdong Basic and Applied Basic Research Foundation(No.2021A1515110221)Fundamental Research Funds for the Central Universities,Sun Yatsen University(No.23yxqntd002)。
文摘Membrane distillation(MD)has gained extensive attention for treating highly saline wastewater.However,membrane scaling during the MD process has hindered the rapid development of this technology.Current approaches to mitigate scaling in membrane distillation focus primarily on achieving enhanced hydrophobicity and even superhydrophobicity via utilizing fluorinated fibrous membrane or introducing perfluorosilane modification.Considering the environmental hazards posed by fluorinated compounds,it is highly desirable to develop non-fluorinated membranes with enhanced anti-scaling properties for effective membrane distillation.In this study,we present a non-fluorinated liquid-like MD membrane with exceptional anti-scaling performance.This membrane was facilely fabricated by grafting linear polydimethylsiloxane(LPDMS)onto a hydrophilic polyether sulfone(PES)membrane pre-coated with the intermediate layers of polydopamine and silica(denoted as LPDMS-PES).Remarkably,LPDMS-PES manifested a drastically improved scaling resistance in continuous MD tests than its perfluorinated counterpart,i.e.,1H,1H,2H,2H-perfluorooctyltrichlorosilane-modified PES membrane(PFOS-PES),in both heterogeneous nucleation-dominated and crystal deposition-dominated scaling processes,despite the latter having a smaller surface energy.LPDMS-PES demonstrated a reduction of crystal accumulation of approximately 85%for Na Cl and 73%for Ca SO_(4) in the heterogeneous nucleation-dominated scaling process compared to PFOS-PES.Additionally,in the crystal deposition-dominated scaling process LPDMS-PES exhibited a reduction of about 70%in scale accumulation.These results explicitly evidenced the great potential of the liquid-like membrane to minimize scaling in membrane distillation by inhibiting both scale nucleation and adhesion onto the membrane.We believe the findings of this study have important implications for the design of high-performance MD membranes,particularly in the quest for environmentally sustainable alternatives to perfluorinated materials.
基金the National Natural Science Foundation of China(22308079)the Natural Science Foundation of Hebei Province,China(B2022202008,B2023202025)the Science and Technology Project of Hebei Education Department,China(BJK2022037).
文摘The operational state of distillation columns significantly impacts product quality and production efficiency.However,due to the complex operation and diverse influencing factors,ensuring the safety and efficient operation of the distillation columns becomes paramount.This research combines passive acoustic monitoring with artificial intelligence techniques,proposed a technology based on residual network(ResNet),which involves the transformation of the acoustic signals emitted by three distillation columns under different operating states.The acoustic signals were initially in one-dimensional waveform format and then converted into two-dimensional Mel-Frequency Cepstral Coefficients spectrogram database using fast Fourier transform.Ultimately,this database was employed to train a ResNet for the purpose of identifying the operational states of the distillation columns.Through this approach,the operational states of distillation columns were monitored.Various faults,including flooding,entrainment,dry-tray,etc.,were diagnosed with an accuracy of 98.91%.Moreover,an intermediate transitional state between normal operation and fault was identified and accurately recognized by the proposed method.Under the transitional state,the acoustic signals achieved an accuracy of 97.85%on the ResNet,which enables early warnings before faults occur,enhancing the safety of chemical production processes.The approach presents a powerful tool for the monitoring and diagnosis of chemical equipment,particularly distillation columns,ensuring the safety and efficiency.
基金National Natural Science Foundation of China(No.62372100)。
文摘To address the challenge of recognizing small target information on traffic panels,a model named MLGIA is proposed based on PaddlePaddle.MLGIA is composed of MobilenetV3 with lightweight GhostBlock(LGB)and an improved augmented feature pyramid network(IAFPN).In this model,LGB improves MobilenetV3 by optimizing the convolutional structure and employing linear transformations to extract sufficient feature maps;IAFPN enhances feature representation through pruning techniques and channel-reduction convolutions.Additionally,knowledge distillation compresses the model and improves its accuracy,while the match category information(MCI)method further optimizes the processing of the detected category information.Experimental results demonstrate that MLGIA outperforms MobilenetV3.MLGIA achieves a detection accuracy comparable to YOLOv8n,with significantly lower resource consumption.Therefore,MLGIA is a strong complement in the traffic panel information recognition domain.
基金supported by the National Natural Science Foundation of China(22378065,22278077 and 22278076)the Key Program of Natural Science Foundation of Fujian Province of China(2022J02019).
文摘The production of high-purity propylene glycol monomethyl ether acetate(PMA)through the transesterification of propylene glycol monomethyl ether(PM)and methyl acetate(MeOAc)is traditionally catalyzed by sodium methoxide.However,the practical application of this method is significantly hindered by the inherent limitations of sodium methoxide,such as its high sensitivity to moisture and propensity for solid precipitation,which impede its effective use in continuous processes.This work proposed a continuous catalytic distillation(CD)process utilizing Amberlyst 15 cation exchange resin as the catalyst.A comprehensive series of reaction kinetic and CD experiments were conducted to evaluate the performance of the proposed process.The results demonstrate that under the optimal operating conditions,namely an ester-to-ether molar ratio of 6:1,a refluxratio of 5:1,a total feed rate of 0.92 g‧min^(-1),and an evaporation rate of 266.47 m^(3)‧m^(-2)‧h^(-1),the conversion rate of PM achieves 99.95%,and the PMA yield is 97.31%.Based on these findings,a process flowsheet for a continuous CD process tailored for the production of electronic-grade PMA is presented.This design incorporates light and heavy removal steps to ensure the production of PMA with a purity of 99.99%.Additionally,the process utilizes pressure swing distillation to recover MeOAc,thereby enhancing the overall efficiencyand sustainability of the production process.The proposed continuous CD process offers a highly efficient,cost-effective,and environmentally sustainable solution for the production of electronic-grade PMA.
基金supported by the National Science Foundation(Grant Nos.NSF-ECCS-2127235 and EFRI-BRAID-2223495)Part of this work was conducted at the Washington Nanofabrication Facility/Molecular Analysis Facility,a National Nanotechnology Coordinated Infrastructure(NNCI)site at the University of Washington with partial support from the National Science Foundation(Grant Nos.NNCI-1542101 and NNCI-2025489).
文摘Optical and hybrid convolutional neural networks(CNNs)recently have become of increasing interest to achieve low-latency,low-power image classification,and computer-vision tasks.However,implementing optical nonlinearity is challenging,and omitting the nonlinear layers in a standard CNN comes with a significant reduction in accuracy.We use knowledge distillation to compress modified AlexNet to a single linear convolutional layer and an electronic backend(two fully connected layers).We obtain comparable performance with a purely electronic CNN with five convolutional layers and three fully connected layers.We implement the convolution optically via engineering the point spread function of an inverse-designed meta-optic.Using this hybrid approach,we estimate a reduction in multiply-accumulate operations from 17M in a conventional electronic modified AlexNet to only 86 K in the hybrid compressed network enabled by the optical front end.This constitutes over 2 orders of magnitude of reduction in latency and power consumption.Furthermore,we experimentally demonstrate that the classification accuracy of the system exceeds 93%on the MNIST dataset of handwritten digits.
基金support from the project grants:Key Research Project in Universities of Henan Province(No.24B480012No.25A450004)Key Specialized Research and Development Breakthrough Program in Henan Province(No.242102240051).
文摘Accurate prediction of perovskite photovoltaic materials’optoelectronic properties is crucial for developing efficient and stable materials,advancing solar technology.To address poor interpretability,high computational complexity,and inaccurate predictions in relevant machine learningmodels,this paper proposes a novelmethodology.The technical route of this papermainly centers on the randomforest-knowledge distillation-bidirectional gated recurrent unit with attention technology(namely RF-KD-BIGRUA),which is applied in perovskite photovoltaic materials.Primarily,it combines random forest to quantitatively assess feature importance,selecting variables with significant impacts on photoelectric conversion efficiency.Subsequently,statistical techniques analyze the weight distribution of variables influencing power conversion efficiency(PCE,%)to extract key features.In the model optimization phase,knowledge distillation transfers features from complex teacher models to student models,enhancing prediction accuracy.Additionally,Bidirectional Gated Recurrent Unit with Attention technology(BiGRU-Attention)is introduced to further optimize predictive performancewhile substantially reducing computational costs.The results demonstrate that integrating statistical techniques into intelligent optimization models can quantify photovoltaic system uncertainties and reduce prediction errors before experimental fabrication,enabling efficient pre-fabrication screening of perovskite materials that meet energy-storage criteria and providing accurate guidance for material selection.
文摘The rapid growth of digital data necessitates advanced natural language processing(NLP)models like BERT(Bidi-rectional Encoder Representations from Transformers),known for its superior performance in text classification.However,BERT’s size and computational demands limit its practicality,especially in resource-constrained settings.This research compresses the BERT base model for Bengali emotion classification through knowledge distillation(KD),pruning,and quantization techniques.Despite Bengali being the sixth most spoken language globally,NLP research in this area is limited.Our approach addresses this gap by creating an efficient BERT-based model for Bengali text.We have explored 20 combinations for KD,quantization,and pruning,resulting in improved speedup,fewer parameters,and reduced memory size.Our best results demonstrate significant improvements in both speed and efficiency.For instance,in the case of mBERT,we achieved a 3.87×speedup and 4×compression ratio with a combination of Distil+Prune+Quant that reduced parameters from 178 to 46 M,while the memory size decreased from 711 to 178 MB.These results offer scalable solutions for NLP tasks in various languages and advance the field of model compression,making these models suitable for real-world applications in resource-limited environments.
基金supported in part by the Jiangsu Province Construction System Science and Technology Project(No.2024ZD056)the Research Development Fund of Xi’an Jiaotong-Liverpool University(No.RDF-24-01-097).
文摘Computer-vision and deep-learning techniques are widely applied to detect,monitor,and assess pavement conditions including road crack detection.Traditional methods fail to achieve satisfactory accuracy and generalization performance in for crack detection.Complex network model can generate redundant feature maps and computational complexity.Therefore,this paper proposes a novel model compression framework based on deep learning to detect road cracks,which can improve the detection efficiency and accuracy.A distillation loss function is proposed to compress the teacher model,followed by channel pruning.Meanwhile,a multi-dilation model is proposed to improve the accuracy of the model pruned.The proposed method is tested on the public database CrackForest dataset(CFD).The experimental results show that the proposed method is more efficient and accurate than other state-of-art methods.
基金Project supported by the National Natural Science Foundation of China(Grant No.62176217)the Program from the Sichuan Provincial Science and Technology,China(Grant No.2018RZ0081)the Fundamental Research Funds of China West Normal University(Grant No.17E063).
文摘Graph neural networks(GNNs)have demonstrated excellent performance in graph representation learning.However,as the volume of graph data grows,issues related to cost and efficiency become increasingly prominent.Graph distillation methods address this challenge by extracting a smaller,reduced graph,ensuring that GNNs trained on both the original and reduced graphs show similar performance.Existing methods,however,primarily optimize the feature matrix of the reduced graph and rely on correlation information from GNNs,while neglecting the original graph’s structure and redundant nodes.This often results in a loss of critical information within the reduced graph.To overcome this limitation,we propose a graph distillation method guided by network symmetry.Specifically,we identify symmetric nodes with equivalent neighborhood structures and merge them into“super nodes”,thereby simplifying the network structure,reducing redundant parameter optimization and enhancing training efficiency.At the same time,instead of relying on the original node features,we employ gradient descent to match optimal features that align with the original features,thus improving downstream task performance.Theoretically,our method guarantees that the reduced graph retains the key information present in the original graph.Extensive experiments demonstrate that our approach achieves significant improvements in graph distillation,exhibiting strong generalization capability and outperforming existing graph reduction methods.
基金supported in part by the National Natural Science Foundation of China under Grants 32171909,52205254,32301704the Guangdong Basic and Applied Basic Research Foundation under Grants 2023A1515011255,2024A1515010199+1 种基金the Scientific Research Projects of Universities in Guangdong Province under Grants 2024ZDZX1042,2024ZDZX3057the Ji-Hua Laboratory Open Project under Grant X220931UZ230.
文摘Defect detection based on computer vision is a critical component in ensuring the quality of industrial products.However,existing detection methods encounter several challenges in practical applications,including the scarcity of labeled samples,limited adaptability of pre-trained models,and the data heterogeneity in distributed environments.To address these issues,this research proposes an unsupervised defect detection method,FLAME(Federated Learning with Adaptive Multi-Model Embeddings).The method comprises three stages:(1)Feature learning stage:this work proposes FADE(Feature-Adaptive Domain-Specific Embeddings),a framework employs Gaussian noise injection to simulate defective patterns and implements a feature discriminator for defect detection,thereby enhancing the pre-trained model’s industrial imagery representation capabilities.(2)Knowledge distillation co-training stage:a multi-model feature knowledge distillation mechanism is introduced.Through feature-level knowledge transfer between the global model and historical local models,the current local model is guided to learn better feature representations from the global model.The approach prevents local models from converging to local optima and mitigates performance degradation caused by data heterogeneity.(3)Model parameter aggregation stage:participating clients utilize weighted averaging aggregation to synthesize an updated global model,facilitating efficient knowledge consolidation.Experimental results demonstrate that FADE improves the average image-level Area under the Receiver Operating Characteristic Curve(AUROC)by 7.34%compared to methods directly utilizing pre-trained models.In federated learning environments,FLAME’s multi-model feature knowledge distillation mechanism outperforms the classic FedAvg algorithm by 2.34%in average image-level AUROC,while exhibiting superior convergence properties.
基金in part by the National Science Foundation of China under Grant No.62276238in part by the National Science Foundation for Distinguished Young Scholars of China under Grant No.62325602in part by the Natural Science Foundation of Henan,China under Grant No.232300421095.
文摘The Heterogeneous Capacitated Vehicle Routing Problem(HCVRP),which involves efficiently routing vehicles with diverse capacities to fulfill various customer demands at minimal cost,poses an NP-hard challenge in combinatorial optimization.Recently,reinforcement learning approaches such as 2D Array Pointer Networks(2D-Ptr)have demonstrated remarkable speed in decision-making by modeling multiple agents’concurrent choices as a sequence of consecutive actions.However,these learning-based models often struggle with generalization,meaning they cannot seamlessly adapt to new scenarios with varying numbers of vehicles or customers without retraining.Inspired by the potential of multi-teacher knowledge distillation to harness diverse knowledge from multiple sources and craft a comprehensive student model,we propose to enhance the generalization capability of 2D-Ptr through Multiple Teacher-forcing Knowledge Distillation(MTKD).We initially train 12 unique 2D-Ptr models under various settings to serve as teacher models.Subsequently,we randomly sample a teacher model and a batch of problem instances,focusing on those where the chosen teacher performed best.This teacher model then solves these instances,generating high-reward action sequences to guide knowledge transfer to the student model.We conduct rigorous evaluations across four distinct datasets,each comprising four HCVRP instances of varying scales.Our empirical findings underscore the proposed method superiority over existing learning-based methods in terms of both computational efficiency and solution quality.
基金funded by National Natural Science Foundation of China(61603245).
文摘Knowledge distillation(KD)is an emerging model compression technique for learning compact object detector models.Previous KD often focused solely on distilling from the logits layer or the feature intermediate layers,which may limit the comprehensive learning of the student network.Additionally,the imbalance between the foreground and background also affects the performance of the model.To address these issues,this paper employs feature-based distillation to enhance the detection performance of the bounding box localization part,and logit-based distillation to improve the detection performance of the category prediction part.Specifically,for the intermediate layer feature distillation,we introduce feature resampling to reduce the risk of the student model merely imitating the teacher model.At the same time,we incorporate a Spatial Attention Mechanism(SAM)to highlight the foreground features learned by the student model.In terms of output layer feature distillation,we divide the traditional distillation targets into target-class objects and non-target-class objects,aiming to improve overall distillation performance.Furthermore,we introduce a one-to-many matching distillation strategy based on Feature Alignment Module(FAM),which further enhances the studentmodel’s feature representation ability,making its feature distribution closer to that of the teacher model,and thus demonstrating superior localization and classification capabilities in object detection tasks.Experimental results demonstrate that our proposedmethodology outperforms conventional distillation techniques in terms of object detecting performance.
基金supported by the Research Foundation of Yunnan Province No.202001BB050043 and 202105AF150011National Natural Science Foundation of China under Grants No.62162065Provincial Foundation for Leaders of Disciplines in Science and Technology No.2019HB121.
文摘In recent years,deep learning has made significant advancements in skin cancer diagnosis.However,most methods prioritize high prediction accuracy without considering the limitations of computational resources,making them impractical for wearable devices.In this case,knowledge distillation has emerged as an effective method,capable of significantly reducing a model’s reliance on computational and storage resources.Nonetheless,previous research suffers from two limitations:1)the student model can only passively receive knowledge from the teacher model,and 2)the teacher model does not effectively model sample relationships during training,potentially hindering the effective transfer of sample relationship-related knowledge during knowledge distillation.To address these issues,we employ two identical student models,each equipped with a sample relationship module.This design ensures that the student models can mutually learn while modeling sample relationships.We conducted extensive experiments on the ISIC 2019 dataset to validate the effectiveness of our method.The results demonstrate that our approach significantly improves the recognition of various types of skin diseases.Compared to state-of-the-art methods,our approach exhibits higher accuracy and better generalization capabilities.
基金the National Natural Science Foundation of China(No.61962032)。
文摘In order to improve the reconstruction accuracy of magnetic resonance imaging(MRI),an accurate natural image compressed sensing(CS)reconstruction network is proposed,which combines the advantages of model-based and deep learning-based CS-MRI methods.In theory,enhancing geometric texture details in linear reconstruction is possible.First,the optimization problem is decomposed into two problems:linear approximation and geometric compensation.Aimed at the problem of image linear approximation,the data consistency module is used to deal with it.Since the processing process will lose texture details,a neural network layer that explicitly combines image and frequency feature representation is proposed,which is named butterfly dilated geometric distillation network.The network introduces the idea of butterfly operation,skillfully integrates the features of image domain and frequency domain,and avoids the loss of texture details when extracting features in a single domain.Finally,a channel feature fusion module is designed by combining channel attention mechanism and dilated convolution.The attention of the channel makes the final output feature map focus on the more important part,thus improving the feature representation ability.The dilated convolution enlarges the receptive field,thereby obtaining more dense image feature data.The experimental results show that the peak signal-to-noise ratio of the network is 5.43 dB,5.24 dB and 3.89 dB higher than that of ISTA-Net+,FISTA and DGDN networks on the brain data set with a Cartesian sampling mask CS ratio of 10%.
