The purpose of this paper is to introduce the notions of m-asymmetric semiopen sets and M-asymmetric semicontinuous multifunctions defined between asymmetric sets satisfying certain minimal conditions in the framework...The purpose of this paper is to introduce the notions of m-asymmetric semiopen sets and M-asymmetric semicontinuous multifunctions defined between asymmetric sets satisfying certain minimal conditions in the framework of bitopological spaces. Some new characterizations of m-asymmetric semiopen sets and M-asymmetric semicontinuous multifunctions will be investigated and several fundamental properties will be obtained.展开更多
In this paper, we provide a new effective method for computing the exact value of Hausdorff measures of a class of self-similar sets satisfying the open set condition （OSC）. As applications, we discuss a self-simila...In this paper, we provide a new effective method for computing the exact value of Hausdorff measures of a class of self-similar sets satisfying the open set condition （OSC）. As applications, we discuss a self-similar Cantor set satisfying OSC and give a simple method for computing its exact Hausdorff measure.展开更多
In this paper, we study the integral solution operators for the -equations on pseudoconvex domains. As a generalization of [1] for the -dequations on pseudoconvex domains with boundary of class C∞, we obtain the ex...In this paper, we study the integral solution operators for the -equations on pseudoconvex domains. As a generalization of [1] for the -dequations on pseudoconvex domains with boundary of class C∞, we obtain the explicit integral operator solutions of C -form for the -equations on pseudoconvex open sets with boundary of Ck (k≥0) and the sup-norm estimates of which solutions have similar as that [1] in form.展开更多
This paper proposes a novel open set recognition method,the Spatial Distribution Feature Extraction Network(SDFEN),to address the problem of electromagnetic signal recognition in an open environment.The spatial distri...This paper proposes a novel open set recognition method,the Spatial Distribution Feature Extraction Network(SDFEN),to address the problem of electromagnetic signal recognition in an open environment.The spatial distribution feature extraction layer in SDFEN replaces convolutional output neural networks with the spatial distribution features that focus more on inter-sample information by incorporating class center vectors.The designed hybrid loss function considers both intra-class distance and inter-class distance,thereby enhancing the similarity among samples of the same class and increasing the dissimilarity between samples of different classes during training.Consequently,this method allows unknown classes to occupy a larger space in the feature space.This reduces the possibility of overlap with known class samples and makes the boundaries between known and unknown samples more distinct.Additionally,the feature comparator threshold can be used to reject unknown samples.For signal open set recognition,seven methods,including the proposed method,are applied to two kinds of electromagnetic signal data:modulation signal and real-world emitter.The experimental results demonstrate that the proposed method outperforms the other six methods overall in a simulated open environment.Specifically,compared to the state-of-the-art Openmax method,the novel method achieves up to 8.87%and 5.25%higher micro-F-measures,respectively.展开更多
We propose and discuss a novel concept of robust set stabilization by permissible controls; this concept is helpful when dealing with both a priori information of model parameters and different permissible controls in...We propose and discuss a novel concept of robust set stabilization by permissible controls; this concept is helpful when dealing with both a priori information of model parameters and different permissible controls including quantum measurements. Both controllability and stabilization can be regarded as the special case of the novel concept. An instance is presented for a kind of uncertain open quantum systems to further justify this gen- eralized concept. It is underlined that a new type of hybrid control based on periodically perturbed projective measurements can be the permissible control of uncertain open quantum systems when perturbed projective measurements are available. The sufficient conditions are given for the robust set stabilization of uncertain quantum open systems by the hybrid control, and the design of the hybrid control is reduced to selecting the period of measurements.展开更多
The new notions of H-metric spaces and generalized H-KKM mappings were introduced. Some generalized H-KKM type theorems for generalized H-K-KM mappings with finitely metrically compactly closed values and finitely met...The new notions of H-metric spaces and generalized H-KKM mappings were introduced. Some generalized H-KKM type theorems for generalized H-K-KM mappings with finitely metrically compactly closed values and finitely metrically compactly open values were established in H-metric spaces. These theorems generalize recent results of Khamsi and Yuan. As applications, some Ky Fan type matching theorems for finitely metrically compactly open covers and finitely metrically compactly closed covers, fixed point theorems and minimax inequality are obtained in H-metric spaces. These results generalize a number of known results in recent literature.展开更多
Let E be a self-similar set satisfying the open set condition. Professor Xu conjectures in his doctoral degree thesis that if H^8（E） 〈｜E｜^8, then for any x ∈ E, the inequality ^-D^3C（E,x）〉H^8（E）/｜E｜^8hold...Let E be a self-similar set satisfying the open set condition. Professor Xu conjectures in his doctoral degree thesis that if H^8（E） 〈｜E｜^8, then for any x ∈ E, the inequality ^-D^3C（E,x）〉H^8（E）/｜E｜^8holds, where 3 = dimH（E）. The above conjecture is negatively answered in this＇paper.展开更多
The neutrality’s origin,character,and extent are studied in the Neutrosophic set.The neutrosophic set is an essential issue to research since it opens the door to a wide range of scientific and technological applicat...The neutrality’s origin,character,and extent are studied in the Neutrosophic set.The neutrosophic set is an essential issue to research since it opens the door to a wide range of scientific and technological applications.The neutrosophic set can find its spot to research because the universe is filled with indeterminacy.Neutrosophic set is currently being developed to express uncertain,imprecise,partial,and inconsistent data.Truth membership function,indeterminacymembership function,and falsitymembership function are used to express a neutrosophic set in order to address uncertainty.The neutrosophic set producesmore rational conclusions in a variety of practical problems.The neutrosophic set displays inconsistencies in data and can solve real-world problems.We are directed to do our work in semi-continuous and almost continuous mapping on the basis of the neutrosophic set by observing these.Since we are going to study the properties of semi continuous and almost continuous mapping,we present the meaning of N-semi-open set,N-semi-closed set,N-regularly open set,N-regularly closed set,N-continuous mapping,N-open mapping,N-closed mapping,Nsemi-continuous mapping,N-semi-open mapping,N-semi-closed mapping.Additionally,we attempt to demonstrate a portion of their properties and furthermore referred to some examples.展开更多
For a self-similar set E satisfying the open set condition,upper convex density is an important concept for the computation of its Hausdorff measure,and it is well known that the set of relative interior points with u...For a self-similar set E satisfying the open set condition,upper convex density is an important concept for the computation of its Hausdorff measure,and it is well known that the set of relative interior points with upper convex density 1 has a full Hausdorff measure.But whether the upper convex densities of E at all the relative interior points are equal to 1？ In other words,whether there exists a relative interior point of E such that the upper convex density of E at this point is less than 1？ In this paper,the authors construct a self-similar set satisfying the open set condition,which has a relative interior point with upper convex density less than 1.Thereby,the above problem is sufficiently answered.展开更多
In this paper,first we define the concept of strong open set for topological space. Then we prove: 1 ) a distributive lattice L possesses the smallest set of generating elements if and only if the Stone space (L) of L...In this paper,first we define the concept of strong open set for topological space. Then we prove: 1 ) a distributive lattice L possesses the smallest set of generating elements if and only if the Stone space (L) of L possesses a smallest base constructed by strong open sets; 2) a Bollean lattice B possesses the smallest set of generating elements if and only if B is a finite Boolean lattice.展开更多
In the present, the authors investigate a new type of separation axioms, which they call it w s-regular. The authors obtained some of its basic properties and its characterizations. Also, the authors notice that the a...In the present, the authors investigate a new type of separation axioms, which they call it w s-regular. The authors obtained some of its basic properties and its characterizations. Also, the authors notice that the axiom of tO s-regularity is weaker than the regularity, stronger than s-regularity and it is independent of w -regularity. However, the authors showed that the w s-regularity and regularity are identical on the class of all locally countable spaces, while the concepts ofw s-regularity and s-regularity are same on the class of anti-locally countable spaces：; furthermore, they proved that the three concepts w s-regularity, s-regularity and w s-regularity are same on the class of extremally disconnected spaces. The authors characterized w s-regular Trspaces by g-open sets, and they proved that the w s-regularity is an open hereditary property and it is also a topologizal property. The w s-closure of subsets of topological spaces are investigated and characterized. The authors used the concepts w s-closure to obtain some characterizations of the w s-regular spaces. Behind those, the authors obtained some properties and characterizations of w -semi open sets.展开更多
In this paper, our focus is to investigate the notion of irresolute topological vector spaces. Irresolute topological vector spaces are defined by using semi open sets and irresolute mappings. The notion of irresolute...In this paper, our focus is to investigate the notion of irresolute topological vector spaces. Irresolute topological vector spaces are defined by using semi open sets and irresolute mappings. The notion of irresolute topological vector spaces is analog to the notion of topological vector spaces, but mathematically it behaves differently. An example is given to show that an irresolute topological vector space is not a topological vector space. It is proved that: 1) Irresolute topological vector spaces possess open hereditary property;2) A homomorphism of irresolute topological vector spaces is irresolute if and only if it is irresolute at identity element;3) In irresolute topological vector spaces, the scalar multiple of semi compact set is semi compact;4) In irresolute topological vector spaces, every semi open set is translationally invariant.展开更多
Recently Lou and Wu obtained the formulas of pointwise dimensions of some Moran measures on Moran sets in Rd under the strong separation condition.In this paper,we prove that the result is still true under the open se...Recently Lou and Wu obtained the formulas of pointwise dimensions of some Moran measures on Moran sets in Rd under the strong separation condition.In this paper,we prove that the result is still true under the open set condition.Due to the lack of the strong separation condition,our approach is essentially different from that used by Lou and Wu.We also obtain the formulas of the Hausdorff and packing dimensions of the Moran measures and discuss some interesting examples.展开更多
We propose to address the open set domain adaptation problem by aligning images at both the pixel space and the feature space.Our approach,called Open Set Translation and Adaptation Network(OSTAN),consists of two main...We propose to address the open set domain adaptation problem by aligning images at both the pixel space and the feature space.Our approach,called Open Set Translation and Adaptation Network(OSTAN),consists of two main components:translation and adaptation.The translation is a cycle-consistent generative adversarial network,which translates any source image to the“style”of a target domain to eliminate domain discrepancy in the pixel space.The adaptation is an instance-weighted adversarial network,which projects both(labeled)translated source images and(unlabeled)target images into a domain-invariant feature space to learn a prior probability for each target image.The learned probability is applied as a weight to the unknown classifier to facilitate the identification of the unknown class.The proposed OSTAN model significantly outperforms the state-of-the-art open set domain adaptation methods on multiple public datasets.Our experiments also demonstrate that both the image-to-image translation and the instance-weighting framework can further improve the decision boundaries for both known and unknown classes.展开更多
In this paper, it is proved that any self-affine set satisfying the strong separation condition is uniformly porous. The author constructs a self-affine set which is not porous, although the open set condition holds. ...In this paper, it is proved that any self-affine set satisfying the strong separation condition is uniformly porous. The author constructs a self-affine set which is not porous, although the open set condition holds. Besides, the author also gives a C^1 iterated function system such that its invariant set is not porous.展开更多
In this paper, an equivalent condition for the self similar sets on the real line to have best coverings is given. As a result, it partly gives answer to the conjecture which was posed by Zhou and Feng [Zhou, Z. L., F...In this paper, an equivalent condition for the self similar sets on the real line to have best coverings is given. As a result, it partly gives answer to the conjecture which was posed by Zhou and Feng [Zhou, Z. L., Feng, L.： Twelve open problems on the exact value of the Hausdorff measure and on topological entropy： A brief survey of recent results. Nonlinearity, 17（2）, 493-502 （2004）].展开更多
Radio frequency fingerprint(RFF)identification is a promising technique for identifying Internet of Things(IoT)devices.This paper presents a comprehensive survey on RFF identification,which covers various aspects rang...Radio frequency fingerprint(RFF)identification is a promising technique for identifying Internet of Things(IoT)devices.This paper presents a comprehensive survey on RFF identification,which covers various aspects ranging from related definitions to details of each stage in the identification process,namely signal preprocessing,RFF feature extraction,further processing,and RFF identification.Specifically,three main steps of preprocessing are summarized,including carrier frequency offset estimation,noise elimination,and channel cancellation.Besides,three kinds of RFFs are categorized,comprising I/Q signal-based,parameter-based,and transformation-based features.Meanwhile,feature fusion and feature dimension reduction are elaborated as two main further processing methods.Furthermore,a novel framework is established from the perspective of closed set and open set problems,and the related state-of-the-art methodologies are investigated,including approaches based on traditional machine learning,deep learning,and generative models.Additionally,we highlight the challenges faced by RFF identification and point out future research trends in this field.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
文摘The purpose of this paper is to introduce the notions of m-asymmetric semiopen sets and M-asymmetric semicontinuous multifunctions defined between asymmetric sets satisfying certain minimal conditions in the framework of bitopological spaces. Some new characterizations of m-asymmetric semiopen sets and M-asymmetric semicontinuous multifunctions will be investigated and several fundamental properties will be obtained.
基金Supported in part by Education Ministry, Anhui province, China (No. KJ2008A028)
文摘In this paper, we provide a new effective method for computing the exact value of Hausdorff measures of a class of self-similar sets satisfying the open set condition （OSC）. As applications, we discuss a self-similar Cantor set satisfying OSC and give a simple method for computing its exact Hausdorff measure.
文摘In this paper, we study the integral solution operators for the -equations on pseudoconvex domains. As a generalization of [1] for the -dequations on pseudoconvex domains with boundary of class C∞, we obtain the explicit integral operator solutions of C -form for the -equations on pseudoconvex open sets with boundary of Ck (k≥0) and the sup-norm estimates of which solutions have similar as that [1] in form.
文摘This paper proposes a novel open set recognition method,the Spatial Distribution Feature Extraction Network(SDFEN),to address the problem of electromagnetic signal recognition in an open environment.The spatial distribution feature extraction layer in SDFEN replaces convolutional output neural networks with the spatial distribution features that focus more on inter-sample information by incorporating class center vectors.The designed hybrid loss function considers both intra-class distance and inter-class distance,thereby enhancing the similarity among samples of the same class and increasing the dissimilarity between samples of different classes during training.Consequently,this method allows unknown classes to occupy a larger space in the feature space.This reduces the possibility of overlap with known class samples and makes the boundaries between known and unknown samples more distinct.Additionally,the feature comparator threshold can be used to reject unknown samples.For signal open set recognition,seven methods,including the proposed method,are applied to two kinds of electromagnetic signal data:modulation signal and real-world emitter.The experimental results demonstrate that the proposed method outperforms the other six methods overall in a simulated open environment.Specifically,compared to the state-of-the-art Openmax method,the novel method achieves up to 8.87%and 5.25%higher micro-F-measures,respectively.
基金Supported by the National Natural Science Foundation of China under Grant Nos 61673389,61273202 and 61134008
文摘We propose and discuss a novel concept of robust set stabilization by permissible controls; this concept is helpful when dealing with both a priori information of model parameters and different permissible controls including quantum measurements. Both controllability and stabilization can be regarded as the special case of the novel concept. An instance is presented for a kind of uncertain open quantum systems to further justify this gen- eralized concept. It is underlined that a new type of hybrid control based on periodically perturbed projective measurements can be the permissible control of uncertain open quantum systems when perturbed projective measurements are available. The sufficient conditions are given for the robust set stabilization of uncertain quantum open systems by the hybrid control, and the design of the hybrid control is reduced to selecting the period of measurements.
文摘The new notions of H-metric spaces and generalized H-KKM mappings were introduced. Some generalized H-KKM type theorems for generalized H-K-KM mappings with finitely metrically compactly closed values and finitely metrically compactly open values were established in H-metric spaces. These theorems generalize recent results of Khamsi and Yuan. As applications, some Ky Fan type matching theorems for finitely metrically compactly open covers and finitely metrically compactly closed covers, fixed point theorems and minimax inequality are obtained in H-metric spaces. These results generalize a number of known results in recent literature.
文摘Let E be a self-similar set satisfying the open set condition. Professor Xu conjectures in his doctoral degree thesis that if H^8（E） 〈｜E｜^8, then for any x ∈ E, the inequality ^-D^3C（E,x）〉H^8（E）/｜E｜^8holds, where 3 = dimH（E）. The above conjecture is negatively answered in this＇paper.
文摘The neutrality’s origin,character,and extent are studied in the Neutrosophic set.The neutrosophic set is an essential issue to research since it opens the door to a wide range of scientific and technological applications.The neutrosophic set can find its spot to research because the universe is filled with indeterminacy.Neutrosophic set is currently being developed to express uncertain,imprecise,partial,and inconsistent data.Truth membership function,indeterminacymembership function,and falsitymembership function are used to express a neutrosophic set in order to address uncertainty.The neutrosophic set producesmore rational conclusions in a variety of practical problems.The neutrosophic set displays inconsistencies in data and can solve real-world problems.We are directed to do our work in semi-continuous and almost continuous mapping on the basis of the neutrosophic set by observing these.Since we are going to study the properties of semi continuous and almost continuous mapping,we present the meaning of N-semi-open set,N-semi-closed set,N-regularly open set,N-regularly closed set,N-continuous mapping,N-open mapping,N-closed mapping,Nsemi-continuous mapping,N-semi-open mapping,N-semi-closed mapping.Additionally,we attempt to demonstrate a portion of their properties and furthermore referred to some examples.
基金The Natural Science Youth Foundation (2008GQS0071) of Jiangxi Province
文摘For a self-similar set E satisfying the open set condition,upper convex density is an important concept for the computation of its Hausdorff measure,and it is well known that the set of relative interior points with upper convex density 1 has a full Hausdorff measure.But whether the upper convex densities of E at all the relative interior points are equal to 1？ In other words,whether there exists a relative interior point of E such that the upper convex density of E at this point is less than 1？ In this paper,the authors construct a self-similar set satisfying the open set condition,which has a relative interior point with upper convex density less than 1.Thereby,the above problem is sufficiently answered.
文摘In this paper,first we define the concept of strong open set for topological space. Then we prove: 1 ) a distributive lattice L possesses the smallest set of generating elements if and only if the Stone space (L) of L possesses a smallest base constructed by strong open sets; 2) a Bollean lattice B possesses the smallest set of generating elements if and only if B is a finite Boolean lattice.
文摘In the present, the authors investigate a new type of separation axioms, which they call it w s-regular. The authors obtained some of its basic properties and its characterizations. Also, the authors notice that the axiom of tO s-regularity is weaker than the regularity, stronger than s-regularity and it is independent of w -regularity. However, the authors showed that the w s-regularity and regularity are identical on the class of all locally countable spaces, while the concepts ofw s-regularity and s-regularity are same on the class of anti-locally countable spaces：; furthermore, they proved that the three concepts w s-regularity, s-regularity and w s-regularity are same on the class of extremally disconnected spaces. The authors characterized w s-regular Trspaces by g-open sets, and they proved that the w s-regularity is an open hereditary property and it is also a topologizal property. The w s-closure of subsets of topological spaces are investigated and characterized. The authors used the concepts w s-closure to obtain some characterizations of the w s-regular spaces. Behind those, the authors obtained some properties and characterizations of w -semi open sets.
文摘In this paper, our focus is to investigate the notion of irresolute topological vector spaces. Irresolute topological vector spaces are defined by using semi open sets and irresolute mappings. The notion of irresolute topological vector spaces is analog to the notion of topological vector spaces, but mathematically it behaves differently. An example is given to show that an irresolute topological vector space is not a topological vector space. It is proved that: 1) Irresolute topological vector spaces possess open hereditary property;2) A homomorphism of irresolute topological vector spaces is irresolute if and only if it is irresolute at identity element;3) In irresolute topological vector spaces, the scalar multiple of semi compact set is semi compact;4) In irresolute topological vector spaces, every semi open set is translationally invariant.
基金supported by National Natural Science Foundation of China (Grant No.11071082)the Fundamental Research Funds for the Central Universities,SCUT
文摘Recently Lou and Wu obtained the formulas of pointwise dimensions of some Moran measures on Moran sets in Rd under the strong separation condition.In this paper,we prove that the result is still true under the open set condition.Due to the lack of the strong separation condition,our approach is essentially different from that used by Lou and Wu.We also obtain the formulas of the Hausdorff and packing dimensions of the Moran measures and discuss some interesting examples.
基金supported by the National Natural Science Foundation of China under Grant Nos.62032011 and 61772257.
文摘We propose to address the open set domain adaptation problem by aligning images at both the pixel space and the feature space.Our approach,called Open Set Translation and Adaptation Network(OSTAN),consists of two main components:translation and adaptation.The translation is a cycle-consistent generative adversarial network,which translates any source image to the“style”of a target domain to eliminate domain discrepancy in the pixel space.The adaptation is an instance-weighted adversarial network,which projects both(labeled)translated source images and(unlabeled)target images into a domain-invariant feature space to learn a prior probability for each target image.The learned probability is applied as a weight to the unknown classifier to facilitate the identification of the unknown class.The proposed OSTAN model significantly outperforms the state-of-the-art open set domain adaptation methods on multiple public datasets.Our experiments also demonstrate that both the image-to-image translation and the instance-weighting framework can further improve the decision boundaries for both known and unknown classes.
基金the National Natural Science Foundation of China(Nos.10671180,10301029,10241003).
文摘In this paper, it is proved that any self-affine set satisfying the strong separation condition is uniformly porous. The author constructs a self-affine set which is not porous, although the open set condition holds. Besides, the author also gives a C^1 iterated function system such that its invariant set is not porous.
基金Supported by National Natural Science Foundations of China (Grant Nos. 10971236, 11261039)
文摘In this paper, an equivalent condition for the self similar sets on the real line to have best coverings is given. As a result, it partly gives answer to the conjecture which was posed by Zhou and Feng [Zhou, Z. L., Feng, L.： Twelve open problems on the exact value of the Hausdorff measure and on topological entropy： A brief survey of recent results. Nonlinearity, 17（2）, 493-502 （2004）].
基金supported in part by the National Natural Science Foundation of China under Grant 62171120 and 62001106National Key Research and Development Program of China(2020YFE0200600)+2 种基金Jiangsu Provincial Key Laboratory of Network and Information Security No.BM2003201Guangdong Key Research and Development Program under Grant2020B0303010001Purple Mountain Laboratories for Network and Communication Security
文摘Radio frequency fingerprint(RFF)identification is a promising technique for identifying Internet of Things(IoT)devices.This paper presents a comprehensive survey on RFF identification,which covers various aspects ranging from related definitions to details of each stage in the identification process,namely signal preprocessing,RFF feature extraction,further processing,and RFF identification.Specifically,three main steps of preprocessing are summarized,including carrier frequency offset estimation,noise elimination,and channel cancellation.Besides,three kinds of RFFs are categorized,comprising I/Q signal-based,parameter-based,and transformation-based features.Meanwhile,feature fusion and feature dimension reduction are elaborated as two main further processing methods.Furthermore,a novel framework is established from the perspective of closed set and open set problems,and the related state-of-the-art methodologies are investigated,including approaches based on traditional machine learning,deep learning,and generative models.Additionally,we highlight the challenges faced by RFF identification and point out future research trends in this field.
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.
