this paper describes a machinelearning algorithm for spoken language acquisition by using concepts extracted from nonliguistic perceptual information, based on a patternrecognition technique. the algorithm projects ...
详细信息
ISBN:
(纸本)0964345692
this paper describes a machinelearning algorithm for spoken language acquisition by using concepts extracted from nonliguistic perceptual information, based on a patternrecognition technique. the algorithm projects the raw sensor-observed signals into perceptually appropriate feature spaces. Lexicon and grammar are respectively represented in stochastic form in the feature space and in the possible grammar parameter space. their learning is based on the association between speech and perceptual information. the syntactic structure is inferred from the conceptual structure obtained by analyzing the conceptual patterns in the perceptual information. the grammar is generalized according to the similarity of concept distributions in the feature space. the algorithm is robust against noise, ambiguity, and sparseness in the learningdata because it uses statistical learning, such as Bayesian learning As a whole, cross-situational learning is implemented in a statistical way. the implemented algorithm includes the processes of the speech recognition and the analysis of graphical scenes containing stationary or moving objects. A preliminary experiment is also described here.
An algorithm for data condensation using support vector machines (SVM's) is presented. the algorithm extracts data points lying close to the class boundaries, which form a much reduced but critical set for classif...
详细信息
In this paper the induction of temporal grammatical rules from multivariate time series is presented in the context of temporal datamining. this includes the use of unsupervised neural networks for the detection of t...
详细信息
ISBN:
(纸本)3540410112
In this paper the induction of temporal grammatical rules from multivariate time series is presented in the context of temporal datamining. this includes the use of unsupervised neural networks for the detection of the most significant temporal patterns in multivariate time series, as well as the use of machinelearning-algorithms for the generation of a rule-based description of primitive patterns. the main idea lies in introducing several abstraction levels for the pattern discovery process. the results of the previous step then are used to induce temporal grammatical rules at different abstraction levels. this approach was successfully applied to a problem in medicine, called sleep apnea.
We present a two-class patternrecognition method through the majority vote which is based on weak classifiers. the weak classifiers are defined in terms of rectangular regions formed by the original training data. Te...
详细信息
We present a two-class patternrecognition method through the majority vote which is based on weak classifiers. the weak classifiers are defined in terms of rectangular regions formed by the original training data. Tests on real and simulated data sets show that this classifier combination procedure can lead to a high accuracy.
In this paper, we investigate the application of support vector machines (SVM) in patternrecognition. SVM is a learning technique developed by Vapnik et al. (1997) that can be seen as a new method for training polyno...
详细信息
In this paper, we investigate the application of support vector machines (SVM) in patternrecognition. SVM is a learning technique developed by Vapnik et al. (1997) that can be seen as a new method for training polynomial, neural network, or radial basis functions classifiers. the decision surfaces are found by solving a linearly constrained quadratic programming problem. We present experimental results of our implementation of SVM, and demonstrate its advantage on well-log data classification problem.
An algorithm for data condensation using support vector machines (SVM) is presented. the algorithm extracts data points lying close to the class boundaries, which form a much reduced but critical set for classificatio...
详细信息
ISBN:
(纸本)0769507506
An algorithm for data condensation using support vector machines (SVM) is presented. the algorithm extracts data points lying close to the class boundaries, which form a much reduced but critical set for classification. the problem of large memory requirements for training SVM in batch mode is circumvented by adopting an active incremental learning algorithm. the learning strategy is motivated from the condensed nearest neighbor classification technique. Experimental results presented show that such active incremental learning enjoy superiority in terms of computation time and condensation ratio, over related methods.
the purpose of this paper is to provide an introductory tutorial on the basic ideas behind support vector machines (SVM). the paper starts with an overview of structural risk minimization (SRM) principle, and describe...
详细信息
the purpose of this paper is to provide an introductory tutorial on the basic ideas behind support vector machines (SVM). the paper starts with an overview of structural risk minimization (SRM) principle, and describes the mechanism of how to construct SVM. For a two-class patternrecognition problem, we discuss in detail the classification mechanism of SVM in three cases of linearly separable, linearly nonseparable and nonlinear. Finally, for nonlinear case, we give a new function mapping technique: By choosing an appropriate kernel function, the SVM can map the low-dimensional input space into the high dimensional feature space, and construct an optimal separating hyperplane with maximum margin in the feature space.
the paper presents the application of three different types of neural networks to the 2D patternrecognition on the basis of its shape. they include the multilayer perceptron (MLP), Kohonen self-organizing network and...
详细信息
the paper presents the application of three different types of neural networks to the 2D patternrecognition on the basis of its shape. they include the multilayer perceptron (MLP), Kohonen self-organizing network and hybrid structure composed of the self-organizing layer and the MLP subnetwork connected in cascade. the recognition is based on the features extracted from the Fourier transform of the data describing the shape of the pattern. Application of different neural network structure results in different accuracy of recognition and classification. the numerical experiments performed for the recognition of the shapes of airplanes have shown the superiority of the hybrid structure.
the recognition of large vocabulary continuous Chinese sign language (CSL) is a challenging problem. It is effective to use phonemes instead of whole signs as the basic units. In this paper, an approach to extracting ...
详细信息
the recognition of large vocabulary continuous Chinese sign language (CSL) is a challenging problem. It is effective to use phonemes instead of whole signs as the basic units. In this paper, an approach to extracting the basic units in CSL automatically is described. In order to find subwords in each data streams in sign signals, dynamic programming (DP) is proposed to segment the data streams, and then ANN approach combining k-means is used to classify these segments. 71 hand postures are automatically extracted from 1063 words and 200 continuous sentences. these postures accompanied with locations and orientations are used as basic units in large vocabulary continuous CSL recognition.
暂无评论