We present a new unsupervised learning technique for the discovery of temporal clusters in large data sets. Our method performs hierarchical decomposition of the data to find structure at many levels of detail and to ...
详细信息
ISBN:
(纸本)0769521223
We present a new unsupervised learning technique for the discovery of temporal clusters in large data sets. Our method performs hierarchical decomposition of the data to find structure at many levels of detail and to reduce the overall computational cost of pattern discovery. We present a comparison to related methods on synthetic data sets and on real gestural and pedestrian flow data.
the C4.5 Decision Tree and Naive Bayes learners are known to produce unreliable probability forecasts. We have used simple Binning [11] and Laplace Transform [2] techniques to improve the reliability of these learners...
详细信息
ISBN:
(纸本)0769521428
the C4.5 Decision Tree and Naive Bayes learners are known to produce unreliable probability forecasts. We have used simple Binning [11] and Laplace Transform [2] techniques to improve the reliability of these learners and compare their effectiveness withthat of the newly developed Venn Probability machine (VPM) meta-learner [9]. We assess improvements in reliability using loss functions, Receiver Operator Characteristic (ROC) curves and Empirical Reliability Curves (ERC). the VPM outperforms the simple techniques to improve reliability, although at the cost of increased computational intensity and slight increase in error rate. these trade-offs are discussed.
We describe how to create withmachinelearning techniques a generative, videorealistic, speech animation module. A human subject is first recorded using a videocamera as he/she utters a pre-determined speech corpus. ...
详细信息
ISBN:
(纸本)0769521223
We describe how to create withmachinelearning techniques a generative, videorealistic, speech animation module. A human subject is first recorded using a videocamera as he/she utters a pre-determined speech corpus. After processing the corpus automatically, a visual speech module is learned from the datathat is capable of synthesizing the human subject's mouth uttering entirely novel utterances that were not recorded in the original video. the synthesized utterance is re-composited onto a background sequence which contains natural head and eye movement. the final output is videorealistic in the sense that it looks like a video camera recording of the subject. At run time, the input to the system can be either real audio sequences or synthetic audio produced by a text-to-speech system, as long as they have been phonetically aligned.
the coal of this work is to evaluate the performance of boosting applied to acoustic model training in various tasks in large vocabulary automatic speech recognition. Specifically, we apply the AdaBoost.M2 algorithm -...
详细信息
ISBN:
(纸本)0889864349
the coal of this work is to evaluate the performance of boosting applied to acoustic model training in various tasks in large vocabulary automatic speech recognition. Specifically, we apply the AdaBoost.M2 algorithm - at the level of utterances - to maximum likelihood and discriminative training of the acoustic parameters of a Hidden Markov Model based speech recognizer. In an isolated word recognition task, boosting improves the best test error rates obtained with discriminative training, even when evaluating final classifiers with a comparable number of parameters. this is demonstrated in a matched and a mismatched decoding task. the second issue of our work is the extension of our algorithm to continuous speech recognition. To this end, we propose an approach realizing the combination of boosted acoustic models at a lexical level, allowing for an online (single-pass) decoding setup for the boosted models. First results are presented for maximum likelihood training in a real-life spontaneous speech dictation task with a 60k word vocabulary and about 58h of training data.
Malicious intrusions (hacking) into computer systems caught the international interest during the recent years. Network administrators are looking for new ways to protect their resources from hackers. there is a stron...
详细信息
ISBN:
(纸本)0889864365
Malicious intrusions (hacking) into computer systems caught the international interest during the recent years. Network administrators are looking for new ways to protect their resources from hackers. there is a strong need for novel strategies for infrastructure protection. the present available techniques are versatile towards misuse detection and difficult to detect the anomalies. Researchers used neural network models, decision trees, statistical models, and rule-based systems with limited success in detecting anomalies. In recent years, research was diverted towards the application of datamining models [10-13, 18] to intrusion detection. More explorations are continuing for new paradigms and programming techniques. Application of genetic algorithm (GA) models is one of the recent explorations. Researchers have better hope with genetic algorithms [7. 17] and bioinformatics [8] applications. In this paper, we select the key attributes from audit data and presented in patterns to compute inductively learned classifiers that,an recognize anomalies and known intrusions. We used the Bucket Bridge algorithm of the genetics based machinelearning to identify the anomalies. Simulation results were presented to detect the anomalies.
Spiking neural networks represent a more plausible model of real biological neurons where time is considered as an important feature for information representation and processing in the human brain. In this paper, we ...
详细信息
ISBN:
(纸本)9812388737
Spiking neural networks represent a more plausible model of real biological neurons where time is considered as an important feature for information representation and processing in the human brain. In this paper, we apply spiking neural networks with dynamic synapses for patternrecognition in multidimensional data. the neurons are based on the integrate and-fire model, and are connected using a biologically plausible model of dynamic synapses. Unlike the conventional synapse employed in artificial neural networks, which is considered as a static entity with a fixed weight, the dynamic synapse (weightless synapse) efficacy changes upon the arrival of input spikes, and depends on the temporal structure of the impinging spike train. the training of the free parameters of the spiking network is performed using an evolutionary strategy (ES) where real values are used to encode the dynamic synapse parameters, which underlie the learning process.. the results show that spiking neurons with dynamic synapses are capable of patternrecognition by means of spatio-temporal encoding.
Examination of advantages and disadvantages of some not commonly used time-frequency representations of vibration signals has been the aim of the paper. the study has been mainly devoted to Wigner-Ville decomposition ...
详细信息
ISBN:
(纸本)0819454362
Examination of advantages and disadvantages of some not commonly used time-frequency representations of vibration signals has been the aim of the paper. the study has been mainly devoted to Wigner-Ville decomposition as well as instantaneous amplitude and frequency. these representations have been examined from the patternrecognition point of view. the Wigner-Ville decomposition was compared with short time Fourier transform taking into account its classification power. In the case of instantaneous amplitude and frequency representation a new method of feature extraction followed by classification using optimal neural classifier has been proposed. Results have been illustrated using a data set of signals measured by a laser vibrometer. they proved that the method proposed in the paper could be used for very fast classification based on vibration signals measured in transient state.
Ensembles of learnt models constitute one of the main current directions in machinelearning and datamining. It was shown experimentally and theoretically that in order for an ensemble to be effective, it should cons...
详细信息
ISBN:
(纸本)354022937X
Ensembles of learnt models constitute one of the main current directions in machinelearning and datamining. It was shown experimentally and theoretically that in order for an ensemble to be effective, it should consist of classifiers having diversity in their predictions. A number of ways are known to quantify diversity in ensembles, but little research has been done about their appropriateness. In this paper, we compare eight measures of the ensemble diversity with regard to their correlation withthe accuracy improvement due to ensembles. We conduct experiments on 21 data sets from the UCI machinelearning repository, comparing the correlations for random subspacing ensembles with different ensemble sizes and with six different ensemble integration methods. Our experiments show that the greatest correlation of the accuracy improvement, on average, is withthe disagreement, entropy, and ambiguity diversity measures, and the lowest correlation, surprisingly, is withthe Q and double fault measures. Normally, the correlation decreases linearly as the ensemble size increases. Much higher correlation values can be seen withthe dynamic integration methods, which are shown to better utilize the ensemble diversity than their static analogues.
the paper describes the "Rough Sets database System" (called in short the RSDS system) for the creation of bibliography on rough sets and their applications. this database is the most comprehensive online ro...
详细信息
ISBN:
(纸本)3540221174
the paper describes the "Rough Sets database System" (called in short the RSDS system) for the creation of bibliography on rough sets and their applications. this database is the most comprehensive online rough sets bibliography and accessible under the following web-site address: http://*** the service has been developed in order to facilitate the creation of rough sets bibliography, for various types of publications. At the moment the bibliography contains over 1400 entries from more than 450 authors. It is possible to create the bibliography in HTML or BibTeX format. In order to broaden the service contents it is possible to append new data using specially dedicated form. After appending data online the database is updated automatically. If one prefers sending a data file to the database administrator, please be aware that the database is updated once a month. In the current version of the RSDS system, there is the possibility for appending to each publication an abstract and keywords. As a natural consequence of this improvement there exists a possibility for searching a publication by keywords.
this paper reports design of a pattern classifying machine (PCM) for distributed datamining (DDM) environment. the proposed PCM is based on the computing model of a special class of sparse network referred to as Cell...
详细信息
暂无评论