Scene graph prediction has emerged as a critical task in computer vision,focusing on transforming complex visual scenes into structured representations by identifying objects,their attributes,and the relationships amo...Scene graph prediction has emerged as a critical task in computer vision,focusing on transforming complex visual scenes into structured representations by identifying objects,their attributes,and the relationships among them.Extending this to 3D semantic scene graph(3DSSG)prediction introduces an additional layer of complexity because it requires the processing of point-cloud data to accurately capture the spatial and volumetric characteristics of a scene.A significant challenge in 3DSSG is the long-tailed distribution of object and relationship labels,causing certain classes to be severely underrepresented and suboptimal performance in these rare categories.To address this,we proposed a fusion prototypical network(FPN),which combines the strengths of conventional neural networks for 3DSSG with a Prototypical Network.The former are known for their ability to handle complex scene graph predictions while the latter excels in few-shot learning scenarios.By leveraging this fusion,our approach enhances the overall prediction accuracy and substantially improves the handling of underrepresented labels.Through extensive experiments using the 3DSSG dataset,we demonstrated that the FPN achieves state-of-the-art performance in 3D scene graph prediction as a single model and effectively mitigates the impact of the long-tailed distribution,providing a more balanced and comprehensive understanding of complex 3D environments.展开更多
In order to improve the models capability in expressing features during few-shot learning,a multi-scale features prototypical network(MS-PN)algorithm is proposed.The metric learning algo-rithm is employed to extract i...In order to improve the models capability in expressing features during few-shot learning,a multi-scale features prototypical network(MS-PN)algorithm is proposed.The metric learning algo-rithm is employed to extract image features and project them into a feature space,thus evaluating the similarity between samples based on their relative distances within the metric space.To sufficiently extract feature information from limited sample data and mitigate the impact of constrained data vol-ume,a multi-scale feature extraction network is presented to capture data features at various scales during the process of image feature extraction.Additionally,the position of the prototype is fine-tuned by assigning weights to data points to mitigate the influence of outliers on the experiment.The loss function integrates contrastive loss and label-smoothing to bring similar data points closer and separate dissimilar data points within the metric space.Experimental evaluations are conducted on small-sample datasets mini-ImageNet and CUB200-2011.The method in this paper can achieve higher classification accuracy.Specifically,in the 5-way 1-shot experiment,classification accuracy reaches 50.13%and 66.79%respectively on these two datasets.Moreover,in the 5-way 5-shot ex-periment,accuracy of 66.79%and 85.91%are observed,respectively.展开更多
Existing almost deep learning methods rely on a large amount of annotated data, so they are inappropriate for forest fire smoke detection with limited data. In this paper, a novel hybrid attention-based few-shot learn...Existing almost deep learning methods rely on a large amount of annotated data, so they are inappropriate for forest fire smoke detection with limited data. In this paper, a novel hybrid attention-based few-shot learning method, named Attention-Based Prototypical Network, is proposed for forest fire smoke detection. Specifically, feature extraction network, which consists of convolutional block attention module, could extract high-level and discriminative features and further decrease the false alarm rate resulting from suspected smoke areas. Moreover, we design a metalearning module to alleviate the overfitting issue caused by limited smoke images, and the meta-learning network enables achieving effective detection via comparing the distance between the class prototype of support images and the features of query images. A series of experiments on forest fire smoke datasets and miniImageNet dataset testify that the proposed method is superior to state-of-the-art few-shot learning approaches.展开更多
Few-shot Learning algorithms can be effectively applied to fields where certain categories have only a small amount of data or a small amount of labeled data,such as medical images,terrorist surveillance,and so on.The...Few-shot Learning algorithms can be effectively applied to fields where certain categories have only a small amount of data or a small amount of labeled data,such as medical images,terrorist surveillance,and so on.The Metric Learning in the Few-shot Learning algorithmis classified by measuring the similarity between the classified samples and the unclassified samples.This paper improves the Prototypical Network in the Metric Learning,and changes its core metric function to Manhattan distance.The Convolutional Neural Network of the embedded module is changed,and mechanisms such as average pooling and Dropout are added.Through comparative experiments,it is found that thismodel can converge in a small number of iterations(below 15,000 episodes),and its performance exceeds algorithms such asMAML.Research shows that replacingManhattan distance with Euclidean distance can effectively improve the classification effect of the Prototypical Network,and mechanisms such as average pooling and Dropout can also effectively improve the model.展开更多
Air quality estimation assesses the pollution level in the air,supports public health warnings,and is a valuable tool in environmental management.Although air sensors have proven helpful in this task,sensors are often...Air quality estimation assesses the pollution level in the air,supports public health warnings,and is a valuable tool in environmental management.Although air sensors have proven helpful in this task,sensors are often expensive and difficult to install,while cameras are becoming more popular and accessible,from which images can be collected as data for deep learning models to solve the above task.This leads to another problem:several labeled images are needed to achieve high accuracy when deep-learningmodels predict air quality.In this research,we have threemain contributions:(1)Collect and publish an air quality estimation dataset,namely PTIT_AQED,including environmental image data and air quality;(2)Propose a deep learning model to predict air quality with few data,called PTIT_FAQE(PTIT Few-shot air quality estimation).We build PTIT_FAQE based on EfficientNet-a CNN architecture that ensures high performance in deep learning applications and Few-shot Learning with Prototypical Networks.This helps the model use only a fewtraining data but still achieve high accuracy in air quality estimation.And(3)conduct experiments to prove the superiority of PTIT_FAQE compared to other studies on both PTIT_AQED and APIN datasets.The results show that our model achieves an accuracy of 0.9278 and an F1-Score of 0.9139 on the PTIT_AQED dataset and an accuracy of 0.9467 and an F1-Score of 0.9371 on the APIN dataset,which demonstrate a significant performance improvement compared to previous studies.We also conduct detailed experiments to evaluate the impact of each component on model performance.展开更多
The few-shot named entity recognition(NER)task aims to train a robust model in the source domain and transfer it to the target domain with very few annotated data.Currently,some approaches rely on the prototypical net...The few-shot named entity recognition(NER)task aims to train a robust model in the source domain and transfer it to the target domain with very few annotated data.Currently,some approaches rely on the prototypical network for NER.However,these approaches often overlook the spatial relations in the span boundary matrix because entity words tend to depend more on adjacent words.We propose using a multidimensional convolution module to address this limitation to capture short-distance spatial dependencies.Additionally,we uti-lize an improved prototypical network and assign different weights to different samples that belong to the same class,thereby enhancing the performance of the few-shot NER task.Further experimental analysis demonstrates that our approach has significantly improved over baseline models across multiple datasets.展开更多
Deep learning significantly improves the accuracy of remote sensing image scene classification,benefiting from the large-scale datasets.However,annotating the remote sensing images is time-consuming and even tough for...Deep learning significantly improves the accuracy of remote sensing image scene classification,benefiting from the large-scale datasets.However,annotating the remote sensing images is time-consuming and even tough for experts.Deep neural networks trained using a few labeled samples usually generalize less to new unseen images.In this paper,we propose a semi-supervised approach for remote sensing image scene classification based on the prototype-based consistency,by exploring massive unlabeled images.To this end,we,first,propose a feature enhancement module to extract discriminative features.This is achieved by focusing the model on the foreground areas.Then,the prototype-based classifier is introduced to the framework,which is used to acquire consistent feature representations.We conduct a series of experiments on NWPU-RESISC45 and Aerial Image Dataset(AID).Our method improves the State-Of-The-Art(SOTA)method on NWPU-RESISC45 from 92.03%to 93.08%and on AID from 94.25%to 95.24%in terms of accuracy.展开更多
Clustering analysis is one of the main concerns in data mining.A common approach to the clustering process is to bring together points that are close to each other and separate points that are away from each other.The...Clustering analysis is one of the main concerns in data mining.A common approach to the clustering process is to bring together points that are close to each other and separate points that are away from each other.Therefore,measuring the distance between sample points is crucial to the effectiveness of clustering.Filtering features by label information and mea-suring the distance between samples by these features is a common supervised learning method to reconstruct distance metric.However,in many application scenarios,it is very expensive to obtain a large number of labeled samples.In this paper,to solve the clustering problem in the few supervised sample and high data dimensionality scenarios,a novel semi-supervised clustering algorithm is proposed by designing an improved prototype network that attempts to reconstruct the distance metric in the sample space with a small amount of pairwise supervised information,such as Must-Link and Cannot-Link,and then cluster the data in the new metric space.The core idea is to make the similar ones closer and the dissimilar ones further away through embedding mapping.Extensive experiments on both real-world and synthetic datasets show the effectiveness of this algorithm.Average clustering metrics on various datasets improved by 8%compared to the comparison algorithm.展开更多
Few-shot named entity recognition(NER)aims to identify named entities in new domains using a limited amount of annotated data.Previous methods divided this task into entity span detection and entity classification,ach...Few-shot named entity recognition(NER)aims to identify named entities in new domains using a limited amount of annotated data.Previous methods divided this task into entity span detection and entity classification,achieving good results.However these methods are limited by the imbalance between the entity and non-entity categories due to the use of sequence labeling for entity span detection.To this end,a point-proto network(PPN)combining pointer and prototypical networks was proposed.Specifically,the pointer network generates the position of entities in sentences in the entity span detection stage.The prototypical network builds semantic prototypes of entity types and classifies entities based on their distance from these prototypes in the entity classification stage.Moreover,the low-rank adaptation(LoRA)fine-tuning method,which involves freezing the pre-trained weights and injecting a trainable decomposition matrix,reduces the parameters that need to be trained and saved.Extensive experiments on the few-shot NER Dataset(Few-NERD)and Cross-Dataset demonstrate the superiority of PPN in this domain.展开更多
Few-shot Named Entity Recognition(NER)systems are designed to identify new categories of entities with a limited number of labeled examples.A major challenge encountered by these systems is overfitting,particularly pr...Few-shot Named Entity Recognition(NER)systems are designed to identify new categories of entities with a limited number of labeled examples.A major challenge encountered by these systems is overfitting,particularly pronounced in comparison to tasks with ample samples.This overfitting predominantly stems from spurious correlations,a consequence of biases inherent in the selection of a small sample set.In response to this challenge,we introduce a novel approach in this paper:a causal intervention-based method for few-shot NER.Building upon the foundational structure of prototypical networks,our method strategically intervenes in the context to obstruct the indirect association between the context and the label.For scenarios restricted to 1-shot,where contextual intervention is not feasible,our method utilizes incremental learning to intervene at the prototype level.This not only counters overfitting but also aids in alleviating catastrophic forgetting.Additionally,to preliminarily classify entity types,we employ entity detection methods for coarse categorization.Considering the distinct characteristics of the source and target domains in few-shot tasks,we introduce sample reweighting to aid in model transfer and generalization.Through rigorous testing across multiple benchmark datasets,our approach consistently sets new state-of-the-art benchmarks,underscoring its efficacy in few-shot NER applications.展开更多
Classifying patterns of known classes and rejecting ambiguous and novel(also called as out-of-distribution(OOD))inputs are involved in open world pattern recognition.Deep neural network models usually excel in closed-...Classifying patterns of known classes and rejecting ambiguous and novel(also called as out-of-distribution(OOD))inputs are involved in open world pattern recognition.Deep neural network models usually excel in closed-set classification while perform poorly in rejecting OOD inputs.To tackle this problem,numerous methods have been designed to perform open set recognition(OSR)or OOD rejection/detection tasks.Previous methods mostly take post-training score transformation or hybrid models to ensure low scores on OOD inputs while separating known classes.In this paper,we attempt to build a unified framework for building open set classifiers for both classification and OOD rejection.We formulate the open set recognition of K-known-class as a(K+1)-class classification problem with model trained on known-class samples only.By decomposing the K-class problem into K one-versus-all(OVA)binary classification tasks and binding some parameters,we show that combining the scores of OVA classifiers can give(K+1)-class posterior probabilities,which enables classification and OOD rejection in a unified framework.To maintain the closed-set classification accuracy of the OVA trained classifier,we propose a hybrid training strategy combining OVA loss and multi-class cross-entropy loss.We implement the OVA framework and hybrid training strategy on the recently proposed convolutional prototype network and prototype classifier on vision transformer(ViT)backbone.Experiments on popular OSR and OOD detection datasets demonstrate that the proposed framework,using a single multi-class classifier,yields competitive performance in closed-set classification,OOD detection,and misclassification detection.The code is available at https://github.com/zhen-cheng121/CPN_OVA_unified.展开更多
Few-shot learning has been proposed and rapidly emerging as a viable means for completing various tasks.Recently,few-shot models have been used for Named Entity Recognition(NER).Prototypical network shows high efficie...Few-shot learning has been proposed and rapidly emerging as a viable means for completing various tasks.Recently,few-shot models have been used for Named Entity Recognition(NER).Prototypical network shows high efficiency on few-shot NER.However,existing prototypical methods only consider the similarity of tokens in query sets and support sets and ignore the semantic similarity among the sentences which contain these entities.We present a novel model,Few-shot Named Entity Recognition with Joint Token and Sentence Awareness(JTSA),to address the issue.The sentence awareness is introduced to probe the semantic similarity among the sentences.The Token awareness is used to explore the similarity of the tokens.To further improve the robustness and results of the model,we adopt the joint learning scheme on the few-shot NER.Experimental results demonstrate that our model outperforms state-of-the-art models on two standard Fewshot NER datasets.展开更多
In this paper,we study cross-domain relation extraction.Since new data mapping to feature spaces always differs from the previously seen data due to a domain shif,few-shot relation extraction often perform poorly.To s...In this paper,we study cross-domain relation extraction.Since new data mapping to feature spaces always differs from the previously seen data due to a domain shif,few-shot relation extraction often perform poorly.To solve the problems caused by cross-domain,we propose a method for combining the pure entity,relation labels and adversarial(PERLA).We first use entities and complete sentences for separate encoding to obtain context-independent entity features.Then,we combine relation labels which are useful for relation extraction to mitigate context noise.We combine adversarial to reduce the noise caused by cross-domain.We conducted experiments on the publicly available cross-domain relation extraction dataset Fewrel 2.o[1]o,and the results show that our approach improves accuracy and has better transferability for better adaptation to cross-domain tasks.展开更多
基金supported by the Glocal University 30 Project Fund of Gyeongsang National University in 2025.
文摘Scene graph prediction has emerged as a critical task in computer vision,focusing on transforming complex visual scenes into structured representations by identifying objects,their attributes,and the relationships among them.Extending this to 3D semantic scene graph(3DSSG)prediction introduces an additional layer of complexity because it requires the processing of point-cloud data to accurately capture the spatial and volumetric characteristics of a scene.A significant challenge in 3DSSG is the long-tailed distribution of object and relationship labels,causing certain classes to be severely underrepresented and suboptimal performance in these rare categories.To address this,we proposed a fusion prototypical network(FPN),which combines the strengths of conventional neural networks for 3DSSG with a Prototypical Network.The former are known for their ability to handle complex scene graph predictions while the latter excels in few-shot learning scenarios.By leveraging this fusion,our approach enhances the overall prediction accuracy and substantially improves the handling of underrepresented labels.Through extensive experiments using the 3DSSG dataset,we demonstrated that the FPN achieves state-of-the-art performance in 3D scene graph prediction as a single model and effectively mitigates the impact of the long-tailed distribution,providing a more balanced and comprehensive understanding of complex 3D environments.
基金the Scientific Research Foundation of Liaoning Provincial Department of Education(No.LJKZ0139)the Program for Liaoning Excellent Talents in University(No.LR15045).
文摘In order to improve the models capability in expressing features during few-shot learning,a multi-scale features prototypical network(MS-PN)algorithm is proposed.The metric learning algo-rithm is employed to extract image features and project them into a feature space,thus evaluating the similarity between samples based on their relative distances within the metric space.To sufficiently extract feature information from limited sample data and mitigate the impact of constrained data vol-ume,a multi-scale feature extraction network is presented to capture data features at various scales during the process of image feature extraction.Additionally,the position of the prototype is fine-tuned by assigning weights to data points to mitigate the influence of outliers on the experiment.The loss function integrates contrastive loss and label-smoothing to bring similar data points closer and separate dissimilar data points within the metric space.Experimental evaluations are conducted on small-sample datasets mini-ImageNet and CUB200-2011.The method in this paper can achieve higher classification accuracy.Specifically,in the 5-way 1-shot experiment,classification accuracy reaches 50.13%and 66.79%respectively on these two datasets.Moreover,in the 5-way 5-shot ex-periment,accuracy of 66.79%and 85.91%are observed,respectively.
基金The work was supported by the National Key R&D Program of China(Grant No.2020YFC1511601)Fundamental Research Funds for the Central Universities(Grant No.2019SHFWLC01).
文摘Existing almost deep learning methods rely on a large amount of annotated data, so they are inappropriate for forest fire smoke detection with limited data. In this paper, a novel hybrid attention-based few-shot learning method, named Attention-Based Prototypical Network, is proposed for forest fire smoke detection. Specifically, feature extraction network, which consists of convolutional block attention module, could extract high-level and discriminative features and further decrease the false alarm rate resulting from suspected smoke areas. Moreover, we design a metalearning module to alleviate the overfitting issue caused by limited smoke images, and the meta-learning network enables achieving effective detection via comparing the distance between the class prototype of support images and the features of query images. A series of experiments on forest fire smoke datasets and miniImageNet dataset testify that the proposed method is superior to state-of-the-art few-shot learning approaches.
文摘Few-shot Learning algorithms can be effectively applied to fields where certain categories have only a small amount of data or a small amount of labeled data,such as medical images,terrorist surveillance,and so on.The Metric Learning in the Few-shot Learning algorithmis classified by measuring the similarity between the classified samples and the unclassified samples.This paper improves the Prototypical Network in the Metric Learning,and changes its core metric function to Manhattan distance.The Convolutional Neural Network of the embedded module is changed,and mechanisms such as average pooling and Dropout are added.Through comparative experiments,it is found that thismodel can converge in a small number of iterations(below 15,000 episodes),and its performance exceeds algorithms such asMAML.Research shows that replacingManhattan distance with Euclidean distance can effectively improve the classification effect of the Prototypical Network,and mechanisms such as average pooling and Dropout can also effectively improve the model.
文摘Air quality estimation assesses the pollution level in the air,supports public health warnings,and is a valuable tool in environmental management.Although air sensors have proven helpful in this task,sensors are often expensive and difficult to install,while cameras are becoming more popular and accessible,from which images can be collected as data for deep learning models to solve the above task.This leads to another problem:several labeled images are needed to achieve high accuracy when deep-learningmodels predict air quality.In this research,we have threemain contributions:(1)Collect and publish an air quality estimation dataset,namely PTIT_AQED,including environmental image data and air quality;(2)Propose a deep learning model to predict air quality with few data,called PTIT_FAQE(PTIT Few-shot air quality estimation).We build PTIT_FAQE based on EfficientNet-a CNN architecture that ensures high performance in deep learning applications and Few-shot Learning with Prototypical Networks.This helps the model use only a fewtraining data but still achieve high accuracy in air quality estimation.And(3)conduct experiments to prove the superiority of PTIT_FAQE compared to other studies on both PTIT_AQED and APIN datasets.The results show that our model achieves an accuracy of 0.9278 and an F1-Score of 0.9139 on the PTIT_AQED dataset and an accuracy of 0.9467 and an F1-Score of 0.9371 on the APIN dataset,which demonstrate a significant performance improvement compared to previous studies.We also conduct detailed experiments to evaluate the impact of each component on model performance.
基金Supported by the Scientific and Technological Innovation 2030-Major Project of New Generation Artificial Intelligence(2020AAA0109300)Science and Technology Commission of Shanghai Municipality(21DZ2203100)2023 Anhui Province Key Research and Development Plan Project-Special Project of Science and Technology Cooperation(2023i11020002)。
文摘The few-shot named entity recognition(NER)task aims to train a robust model in the source domain and transfer it to the target domain with very few annotated data.Currently,some approaches rely on the prototypical network for NER.However,these approaches often overlook the spatial relations in the span boundary matrix because entity words tend to depend more on adjacent words.We propose using a multidimensional convolution module to address this limitation to capture short-distance spatial dependencies.Additionally,we uti-lize an improved prototypical network and assign different weights to different samples that belong to the same class,thereby enhancing the performance of the few-shot NER task.Further experimental analysis demonstrates that our approach has significantly improved over baseline models across multiple datasets.
基金supported in part by the National Natural Science Foundation of China(No.12302252)。
文摘Deep learning significantly improves the accuracy of remote sensing image scene classification,benefiting from the large-scale datasets.However,annotating the remote sensing images is time-consuming and even tough for experts.Deep neural networks trained using a few labeled samples usually generalize less to new unseen images.In this paper,we propose a semi-supervised approach for remote sensing image scene classification based on the prototype-based consistency,by exploring massive unlabeled images.To this end,we,first,propose a feature enhancement module to extract discriminative features.This is achieved by focusing the model on the foreground areas.Then,the prototype-based classifier is introduced to the framework,which is used to acquire consistent feature representations.We conduct a series of experiments on NWPU-RESISC45 and Aerial Image Dataset(AID).Our method improves the State-Of-The-Art(SOTA)method on NWPU-RESISC45 from 92.03%to 93.08%and on AID from 94.25%to 95.24%in terms of accuracy.
文摘Clustering analysis is one of the main concerns in data mining.A common approach to the clustering process is to bring together points that are close to each other and separate points that are away from each other.Therefore,measuring the distance between sample points is crucial to the effectiveness of clustering.Filtering features by label information and mea-suring the distance between samples by these features is a common supervised learning method to reconstruct distance metric.However,in many application scenarios,it is very expensive to obtain a large number of labeled samples.In this paper,to solve the clustering problem in the few supervised sample and high data dimensionality scenarios,a novel semi-supervised clustering algorithm is proposed by designing an improved prototype network that attempts to reconstruct the distance metric in the sample space with a small amount of pairwise supervised information,such as Must-Link and Cannot-Link,and then cluster the data in the new metric space.The core idea is to make the similar ones closer and the dissimilar ones further away through embedding mapping.Extensive experiments on both real-world and synthetic datasets show the effectiveness of this algorithm.Average clustering metrics on various datasets improved by 8%compared to the comparison algorithm.
基金supported by the National Key Research and Development Project(2021YFF0901701)。
文摘Few-shot named entity recognition(NER)aims to identify named entities in new domains using a limited amount of annotated data.Previous methods divided this task into entity span detection and entity classification,achieving good results.However these methods are limited by the imbalance between the entity and non-entity categories due to the use of sequence labeling for entity span detection.To this end,a point-proto network(PPN)combining pointer and prototypical networks was proposed.Specifically,the pointer network generates the position of entities in sentences in the entity span detection stage.The prototypical network builds semantic prototypes of entity types and classifies entities based on their distance from these prototypes in the entity classification stage.Moreover,the low-rank adaptation(LoRA)fine-tuning method,which involves freezing the pre-trained weights and injecting a trainable decomposition matrix,reduces the parameters that need to be trained and saved.Extensive experiments on the few-shot NER Dataset(Few-NERD)and Cross-Dataset demonstrate the superiority of PPN in this domain.
基金supported by the State Key Program of National Natural Science of China(No.61533018)the National Natural Science Foundation of China(No.61402220)+2 种基金the Philosophy and Social Science Foundation of Hunan Province(No.16YBA323)the Natural Science Foundation of Hunan Province(Nos.2020JJ4525 and 2022JJ30495)the Scientific Research Fund of Hunan Provincial Education Department(Nos.18B279,19A439,and 22A0316).
文摘Few-shot Named Entity Recognition(NER)systems are designed to identify new categories of entities with a limited number of labeled examples.A major challenge encountered by these systems is overfitting,particularly pronounced in comparison to tasks with ample samples.This overfitting predominantly stems from spurious correlations,a consequence of biases inherent in the selection of a small sample set.In response to this challenge,we introduce a novel approach in this paper:a causal intervention-based method for few-shot NER.Building upon the foundational structure of prototypical networks,our method strategically intervenes in the context to obstruct the indirect association between the context and the label.For scenarios restricted to 1-shot,where contextual intervention is not feasible,our method utilizes incremental learning to intervene at the prototype level.This not only counters overfitting but also aids in alleviating catastrophic forgetting.Additionally,to preliminarily classify entity types,we employ entity detection methods for coarse categorization.Considering the distinct characteristics of the source and target domains in few-shot tasks,we introduce sample reweighting to aid in model transfer and generalization.Through rigorous testing across multiple benchmark datasets,our approach consistently sets new state-of-the-art benchmarks,underscoring its efficacy in few-shot NER applications.
基金supported by the National Key Research and Development Program,China(No.2018 AAA0100400)National Natural Science Foundation of China(Nos.U20A20223,62222609 and 62076236).
文摘Classifying patterns of known classes and rejecting ambiguous and novel(also called as out-of-distribution(OOD))inputs are involved in open world pattern recognition.Deep neural network models usually excel in closed-set classification while perform poorly in rejecting OOD inputs.To tackle this problem,numerous methods have been designed to perform open set recognition(OSR)or OOD rejection/detection tasks.Previous methods mostly take post-training score transformation or hybrid models to ensure low scores on OOD inputs while separating known classes.In this paper,we attempt to build a unified framework for building open set classifiers for both classification and OOD rejection.We formulate the open set recognition of K-known-class as a(K+1)-class classification problem with model trained on known-class samples only.By decomposing the K-class problem into K one-versus-all(OVA)binary classification tasks and binding some parameters,we show that combining the scores of OVA classifiers can give(K+1)-class posterior probabilities,which enables classification and OOD rejection in a unified framework.To maintain the closed-set classification accuracy of the OVA trained classifier,we propose a hybrid training strategy combining OVA loss and multi-class cross-entropy loss.We implement the OVA framework and hybrid training strategy on the recently proposed convolutional prototype network and prototype classifier on vision transformer(ViT)backbone.Experiments on popular OSR and OOD detection datasets demonstrate that the proposed framework,using a single multi-class classifier,yields competitive performance in closed-set classification,OOD detection,and misclassification detection.The code is available at https://github.com/zhen-cheng121/CPN_OVA_unified.
基金The State Key Program of National Natural Science of China,Grant/Award Number:61533018National Natural Science Foundation of China,Grant/Award Number:61402220+2 种基金The Philosophy and Social Science Foundation of Hunan Province,Grant/Award Number:16YBA323Natural Science Foundation of Hunan Province,Grant/Award Number:2020J4525,2022JJ30495Scientific Research Fund of Hunan Provincial Education Department,Grant/Award Number:18B279,19A439,22A0316.
文摘Few-shot learning has been proposed and rapidly emerging as a viable means for completing various tasks.Recently,few-shot models have been used for Named Entity Recognition(NER).Prototypical network shows high efficiency on few-shot NER.However,existing prototypical methods only consider the similarity of tokens in query sets and support sets and ignore the semantic similarity among the sentences which contain these entities.We present a novel model,Few-shot Named Entity Recognition with Joint Token and Sentence Awareness(JTSA),to address the issue.The sentence awareness is introduced to probe the semantic similarity among the sentences.The Token awareness is used to explore the similarity of the tokens.To further improve the robustness and results of the model,we adopt the joint learning scheme on the few-shot NER.Experimental results demonstrate that our model outperforms state-of-the-art models on two standard Fewshot NER datasets.
基金The State Key Program of National Natural Science of China,Grant/Award Number:61533018National Natural Science Foundation of China,Grant/Award Number:61402220+2 种基金The Philosophy and Social Science Foundation of Hunan Province,Grant/Award Number:16YBA323Natural Science Foundation of Hunan Province,Grant/Award Number:2020J4525,2022J30495Scientific Research Fund of Hunan Provincial Education Department,Grant/Award Number:18B279,19A439.
文摘In this paper,we study cross-domain relation extraction.Since new data mapping to feature spaces always differs from the previously seen data due to a domain shif,few-shot relation extraction often perform poorly.To solve the problems caused by cross-domain,we propose a method for combining the pure entity,relation labels and adversarial(PERLA).We first use entities and complete sentences for separate encoding to obtain context-independent entity features.Then,we combine relation labels which are useful for relation extraction to mitigate context noise.We combine adversarial to reduce the noise caused by cross-domain.We conducted experiments on the publicly available cross-domain relation extraction dataset Fewrel 2.o[1]o,and the results show that our approach improves accuracy and has better transferability for better adaptation to cross-domain tasks.
