Pub Date : 1999-11-16DOI: 10.1109/ICONIP.1999.843958
D. Merkl, A. Rauber
The self-organizing map is a very popular unsupervised neural network model for the analysis of high-dimensional input data as in information retrieval applications. However, the interpretation of the map requires much manual effort, especially as far as the analysis of the learned features and the characteristics of identified clusters is concerned. We present our novel LabelSOM method which, based on the features learned by the map, automatically selects the most descriptive features of the input patterns mapped onto a particular unit of the map, thus making the characteristics of the various clusters within the map explicit. We demonstrate the benefits of this approach on an example from text classification using a real-world document archive. In this particular case, the features correspond to keywords describing the contents of a document. The benefit of this approach is that the various document clusters are characterized in terms of shared keywords, thus making it easy for the user to explore the contents of an unknown document archive.
{"title":"Automatic labeling of self-organizing maps for information retrieval","authors":"D. Merkl, A. Rauber","doi":"10.1109/ICONIP.1999.843958","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.843958","url":null,"abstract":"The self-organizing map is a very popular unsupervised neural network model for the analysis of high-dimensional input data as in information retrieval applications. However, the interpretation of the map requires much manual effort, especially as far as the analysis of the learned features and the characteristics of identified clusters is concerned. We present our novel LabelSOM method which, based on the features learned by the map, automatically selects the most descriptive features of the input patterns mapped onto a particular unit of the map, thus making the characteristics of the various clusters within the map explicit. We demonstrate the benefits of this approach on an example from text classification using a real-world document archive. In this particular case, the features correspond to keywords describing the contents of a document. The benefit of this approach is that the various document clusters are characterized in terms of shared keywords, thus making it easy for the user to explore the contents of an unknown document archive.","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"17 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128161514","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1999-11-16DOI: 10.1109/ICONIP.1999.845679
M. L. Vaughn, S. Cavill, S. Taylor, M. Foy, A. Fogg
Using a new method designed by the first author, this paper shows how direct explanations in the form of a ranked data relationship can be provided to explain the classification of an input case by a standard multilayer perceptron (MLP) network. It is also shown how knowledge in the form of an induced rule can be discovered from the data relationship for each training case. The method is demonstrated for example training cases from a real-world MLP that classifies low back pain patients into three diagnostic classes. It is shown how the validation of the explanations for all training cases provides a way of validating the low back pain MLP network. In validating the network, a number of test cases apparently mis-classified by the MLP were found to have been correctly classified by the network and incorrectly classified by the clinicians.
{"title":"Using direct explanations to validate a multi-layer perceptron network that classifies low back pain patients","authors":"M. L. Vaughn, S. Cavill, S. Taylor, M. Foy, A. Fogg","doi":"10.1109/ICONIP.1999.845679","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.845679","url":null,"abstract":"Using a new method designed by the first author, this paper shows how direct explanations in the form of a ranked data relationship can be provided to explain the classification of an input case by a standard multilayer perceptron (MLP) network. It is also shown how knowledge in the form of an induced rule can be discovered from the data relationship for each training case. The method is demonstrated for example training cases from a real-world MLP that classifies low back pain patients into three diagnostic classes. It is shown how the validation of the explanations for all training cases provides a way of validating the low back pain MLP network. In validating the network, a number of test cases apparently mis-classified by the MLP were found to have been correctly classified by the network and incorrectly classified by the clinicians.","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"205 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123053560","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1999-11-16DOI: 10.1109/ICONIP.1999.844008
S. Wu, H. Nakahara
We consider two methods to optimize the distribution of preferred stimulus in a population code based on the knowledge of the distribution of stimulus. One method is to maximize the mean Fisher information of the population with respect to the stimulus ensemble. The other is to minimize the lower bound of the mean decoding error. The implication of the two methods is discussed.
{"title":"Optimize the distribution of preferred stimulus in a population code","authors":"S. Wu, H. Nakahara","doi":"10.1109/ICONIP.1999.844008","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.844008","url":null,"abstract":"We consider two methods to optimize the distribution of preferred stimulus in a population code based on the knowledge of the distribution of stimulus. One method is to maximize the mean Fisher information of the population with respect to the stimulus ensemble. The other is to minimize the lower bound of the mean decoding error. The implication of the two methods is discussed.","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"68 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124034421","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1999-11-16DOI: 10.1109/ICONIP.1999.843965
Tom Gedeon, L. Coward, Bai-ling Zhang
Functionally complex electronic systems are organized into functional components exchanging unambiguous information. The requirement to exchange unambiguous information results in difficulties in implementing parallel processing and extreme difficulty in implementing any capability to heuristically change functionality based on experience. The recommendation architecture allows the exchange of ambiguous information between functional components and therefore offers a way to reduce these difficulties. A system with the recommendation architecture uses a device imprinting mechanism to heuristically organize its inputs into a portfolio of ambiguous information repetition conditions on a range of levels of detail. The presence or absence of these conditions contains enough information to be used by a separate subsystem to determine appropriate behavior. Simulations of a simple system with the recommendation architecture demonstrate that sequences of inputs of wide range of different types can be heuristically organized into a functionally usable set of repetition conditions. Organization is successful even though there are no exact repetitions of input conditions. Learning effectiveness measures which make no use of information on the consequences of system actions can be used to adjust architectural parameters to organize even wider ranges of input types. These results demonstrate the feasibility of developing functionally complex systems with the recommendation architecture.
{"title":"Results of simulations of a system with the recommendation architecture","authors":"Tom Gedeon, L. Coward, Bai-ling Zhang","doi":"10.1109/ICONIP.1999.843965","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.843965","url":null,"abstract":"Functionally complex electronic systems are organized into functional components exchanging unambiguous information. The requirement to exchange unambiguous information results in difficulties in implementing parallel processing and extreme difficulty in implementing any capability to heuristically change functionality based on experience. The recommendation architecture allows the exchange of ambiguous information between functional components and therefore offers a way to reduce these difficulties. A system with the recommendation architecture uses a device imprinting mechanism to heuristically organize its inputs into a portfolio of ambiguous information repetition conditions on a range of levels of detail. The presence or absence of these conditions contains enough information to be used by a separate subsystem to determine appropriate behavior. Simulations of a simple system with the recommendation architecture demonstrate that sequences of inputs of wide range of different types can be heuristically organized into a functionally usable set of repetition conditions. Organization is successful even though there are no exact repetitions of input conditions. Learning effectiveness measures which make no use of information on the consequences of system actions can be used to adjust architectural parameters to organize even wider ranges of input types. These results demonstrate the feasibility of developing functionally complex systems with the recommendation architecture.","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"9 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124537265","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1999-11-16DOI: 10.1109/ICONIP.1999.844025
S. Cho, S. Choi, P. Wong
Outliers, noise and data density imbalance, present in most real world data, render it difficult to properly train neural networks. Conventionally residual analysis was used to detect outliers. When used with neural networks, however, the procedure is computationally costly. The authors propose an efficient heuristic data selection method that is based on Bayesian error bars. After a neural network is trained, the residual and error bar are computed for each data. The data that correspond to large residual or large error bars are removed from the training data set. The remaining data are then used to further train the network. The proposed approach was applied to two real world problems: rock porosity and permeability prediction problems in reservoir engineering, with a significant generalization performance improvement of 30-55%. This preliminary result suggests that the approach deserves further investigation.
{"title":"Data selection based on Bayesian error bar","authors":"S. Cho, S. Choi, P. Wong","doi":"10.1109/ICONIP.1999.844025","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.844025","url":null,"abstract":"Outliers, noise and data density imbalance, present in most real world data, render it difficult to properly train neural networks. Conventionally residual analysis was used to detect outliers. When used with neural networks, however, the procedure is computationally costly. The authors propose an efficient heuristic data selection method that is based on Bayesian error bars. After a neural network is trained, the residual and error bar are computed for each data. The data that correspond to large residual or large error bars are removed from the training data set. The remaining data are then used to further train the network. The proposed approach was applied to two real world problems: rock porosity and permeability prediction problems in reservoir engineering, with a significant generalization performance improvement of 30-55%. This preliminary result suggests that the approach deserves further investigation.","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"33 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127662401","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1999-11-16DOI: 10.1109/ICONIP.1999.844710
Daw-Tung Lin, Jing Chen
Proposes a hierarchical model of a radial basis function network to classify and to recognize facial expressions. This approach utilizes principal component analysis as the feature extraction process from static images. It decomposes the acquired data into a small set of characteristic features. Using hierarchical networks of Gaussian radial basis functions, we differentiate the images in the feature space and fulfil the classification task. The objective of this research is to develop a more efficient system to discriminate between seven facial expressions (happiness, sadness, surprise, fear, anger, disgust and neutral). A constructive procedure is detailed and the system performance is evaluated. We achieved a correct classification rate above 98.4%, which is overwhelming distinguished compared to other approaches.
{"title":"Facial expressions classification with hierarchical radial basis function networks","authors":"Daw-Tung Lin, Jing Chen","doi":"10.1109/ICONIP.1999.844710","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.844710","url":null,"abstract":"Proposes a hierarchical model of a radial basis function network to classify and to recognize facial expressions. This approach utilizes principal component analysis as the feature extraction process from static images. It decomposes the acquired data into a small set of characteristic features. Using hierarchical networks of Gaussian radial basis functions, we differentiate the images in the feature space and fulfil the classification task. The objective of this research is to develop a more efficient system to discriminate between seven facial expressions (happiness, sadness, surprise, fear, anger, disgust and neutral). A constructive procedure is detailed and the system performance is evaluated. We achieved a correct classification rate above 98.4%, which is overwhelming distinguished compared to other approaches.","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"78 1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128670283","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1999-11-16DOI: 10.1109/ICONIP.1999.845697
Y. Wada, N. Shimodate
We have shown that a complex motion of the arm can be generated based on the optimization principle of smoothness in which two or more via-points are assumed to be a boundary condition. We have previously proposed a perception model for cursive-connected characters which has these via-points as features (Y. Wada and M. Kawato, 1995). Via-points are representative forms in the computational trajectory formation model of the human arm. The paper shows that a formation conversion from an intention to a set of via-points and a perception conversion from a set of via-points to an intention can be achieved using the same structural recurrent neural network based on bi-directional theory. As a concrete example, we demonstrate the formation and the perception of human gestures. In other words, the model is achieved by applying the motor theory of pattern perception, which is based on bi-directionals using neural networks. Finally, the paper shows that segmentation of a continuous motion is possible, a concept that can be useful to the field of engineering.
{"title":"Neural networks of formation and perception using motion via-points: an application to hand gestures","authors":"Y. Wada, N. Shimodate","doi":"10.1109/ICONIP.1999.845697","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.845697","url":null,"abstract":"We have shown that a complex motion of the arm can be generated based on the optimization principle of smoothness in which two or more via-points are assumed to be a boundary condition. We have previously proposed a perception model for cursive-connected characters which has these via-points as features (Y. Wada and M. Kawato, 1995). Via-points are representative forms in the computational trajectory formation model of the human arm. The paper shows that a formation conversion from an intention to a set of via-points and a perception conversion from a set of via-points to an intention can be achieved using the same structural recurrent neural network based on bi-directional theory. As a concrete example, we demonstrate the formation and the perception of human gestures. In other words, the model is achieved by applying the motor theory of pattern perception, which is based on bi-directionals using neural networks. Finally, the paper shows that segmentation of a continuous motion is possible, a concept that can be useful to the field of engineering.","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"75 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128454836","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1999-11-16DOI: 10.1109/ICONIP.1999.844004
S. Glankwamdee, P. Chariyavilaskul
In this project, a fuzzy supervisory controller system which includes a PID controller, a feedforward controller and a decoupler controller is proposed for a binary distillation column in order to reject the feedflow disturbances affecting the compositions. The scheme uses fuzzy rules and reasoning online in order to determine the controller parameters based on the error signal and its first difference. Simulations show that the control system performs satisfactorily.
{"title":"Fuzzy supervisory control system of a binary distillation column","authors":"S. Glankwamdee, P. Chariyavilaskul","doi":"10.1109/ICONIP.1999.844004","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.844004","url":null,"abstract":"In this project, a fuzzy supervisory controller system which includes a PID controller, a feedforward controller and a decoupler controller is proposed for a binary distillation column in order to reject the feedflow disturbances affecting the compositions. The scheme uses fuzzy rules and reasoning online in order to determine the controller parameters based on the error signal and its first difference. Simulations show that the control system performs satisfactorily.","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115541585","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1999-11-16DOI: 10.1109/ICONIP.1999.844010
A. Labbi, H. Bosch, C. Pellegrini, W. Gerstner
This paper addresses the problem of image categorization using local sensory information which is aggregated into global cortical-like representations of different image categories. Local information is adaptively extracted from an image database using independent component analysis (ICA) which provides a set of localized, oriented, and band-pass filters selective to the most independent features of the different categories. Such local representations have been computationally investigated by several researchers, and have also been experimentally observed as characteristics of simple cell receptive fields in the primary visual cortex. However, little work has been done on further use of these representations to provide more complex and global description of images. In this paper, we present an algorithm which uses the energy of a minimal set of filters to provide category-specific signatures which are shown to be strongly discriminant. Computer simulations are carried on an image database consisting of three categories (faces, leaves, and buildings). The categorization performances of the algorithm using ICA and PCA filters are reported. It is mainly shown that considering a small number of PCA filters leads to a performance which is not significantly improved by considering other PCA filters, however, considering additional ICA filters increases performance due to the fact that each additional filter carries additional information (in the entropy sense).
{"title":"Sparse-distributed codes for image categorization","authors":"A. Labbi, H. Bosch, C. Pellegrini, W. Gerstner","doi":"10.1109/ICONIP.1999.844010","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.844010","url":null,"abstract":"This paper addresses the problem of image categorization using local sensory information which is aggregated into global cortical-like representations of different image categories. Local information is adaptively extracted from an image database using independent component analysis (ICA) which provides a set of localized, oriented, and band-pass filters selective to the most independent features of the different categories. Such local representations have been computationally investigated by several researchers, and have also been experimentally observed as characteristics of simple cell receptive fields in the primary visual cortex. However, little work has been done on further use of these representations to provide more complex and global description of images. In this paper, we present an algorithm which uses the energy of a minimal set of filters to provide category-specific signatures which are shown to be strongly discriminant. Computer simulations are carried on an image database consisting of three categories (faces, leaves, and buildings). The categorization performances of the algorithm using ICA and PCA filters are reported. It is mainly shown that considering a small number of PCA filters leads to a performance which is not significantly improved by considering other PCA filters, however, considering additional ICA filters increases performance due to the fact that each additional filter carries additional information (in the entropy sense).","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"41 4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115136867","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1999-11-16DOI: 10.1109/ICONIP.1999.844666
J. Lu, M. Quaddus, R. Williams
The paper presents a conceptual framework that extends the use of multiple objective decision making (MODM) technique within the knowledge based decision support system architecture. The system can guide users systematically towards the selection and application of the most appropriate method for their decision making. The conceptual framework has been implemented as an intelligent and graphical user interface (GUI) based multiple objective decision support systems prototype, called intelligent multiple objective decision support system (IMODSS).
{"title":"Enhancing a multi-objective decision support system through knowledge-based guidance: a conceptual framework and prototype","authors":"J. Lu, M. Quaddus, R. Williams","doi":"10.1109/ICONIP.1999.844666","DOIUrl":"https://doi.org/10.1109/ICONIP.1999.844666","url":null,"abstract":"The paper presents a conceptual framework that extends the use of multiple objective decision making (MODM) technique within the knowledge based decision support system architecture. The system can guide users systematically towards the selection and application of the most appropriate method for their decision making. The conceptual framework has been implemented as an intelligent and graphical user interface (GUI) based multiple objective decision support systems prototype, called intelligent multiple objective decision support system (IMODSS).","PeriodicalId":237855,"journal":{"name":"ICONIP'99. ANZIIS'99 & ANNES'99 & ACNN'99. 6th International Conference on Neural Information Processing. Proceedings (Cat. No.99EX378)","volume":"178 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1999-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115443560","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：