Jari Kangas

Engineering applications of the self-organizing map

The self-organizing map (SOM) method is a new, powerful software tool for the visualization of high-dimensional data. It converts complex, nonlinear statistical relationships between high-dimensional data into simple geometric relationships on a low-dimensional display. As it thereby compresses information while preserving the most important topological and metric relationships of the primary data elements on the display, it may also be thought to produce some kind of abstractions. The term self-organizing map signifies a class of mappings defined by error-theoretic considerations. In practice they result in certain unsupervised, competitive learning processes, computed by simple-looking SOM algorithms. Many industries have found the SOM-based software tools useful. The most important property of the SOM, orderliness of the input-output mapping, can be utilized for many tasks: reduction of the amount of training data, speeding up learning nonlinear interpolation and extrapolation, generalization, and effective compression of information for its transmission.

Variants of self-organizing maps

Self-organizing maps have a connection with traditional vector quantization. A characteristic which makes them resemble certain biological brain maps, however, is the spatial order of their responses which is formed in the learning process. Two innovations are discussed: dynamic weighting of the input signals at each input of each cell, which improves the ordering when very different input signals are used, and definition of neighborhoods in the learning algorithm by the minimum spanning tree, which provides a far better and faster approximation of prominently structured density functions. It is cautioned that if the maps are used for pattern recognition and decision processes, it is necessary to fine-tune the reference vectors such that they directly define the decision borders.<<ETX>>

LVQPAK: A software package for the correct application of Learning Vector Quantization algorithms

An overview of the software package LVQPAK, which has been developed for convenient and effective application of learning vector quantization algorithms, is presented. Two new features are included: fast conflict-free initial distribution of codebook vectors into the class zones and the optimized-learning-rate algorithm OLVQ1.<<ETX>>

Process State Monitoring Using Self-Organizing Maps

Self-organizing map algorithm has the ability to create a model for a system that is not exactly known a priori. Using this kind of model we can classify the system states and detect the abnormal ones. In this paper, an application of the feature map to detect operational states of a device is presented. The features used in the map are the measurements from the device describing its operational and environmental parameters.

Developments and applications of the self-organizing map and related algorithms

In this paper the basic principles and developments of an unsupervised learning algorithm, the self-organizing map (SOM) and a supervised learning algorithm, the learning vector quantization (LVQ) are explained. Some practical applications of the algorithms in data analysis, data visualization and pattern recognition tasks are mentioned. At the end of the paper new results are reported about increased error tolerance in the transmission of vector quantized images, provided by the topological ordering of codewords by the SOM algorithm.

Time-delayed self-organizing maps

Three related possibilities for representing the sequential aspect of data using the self-organizing map model are studied. The quantitative results of experiments with artificial test data are described, and the most promising solutions for future work are discussed. In the first model, the backwards exponentially averaged input vectors are used as the pattern vector. In the second model, a concatenation model where actual input patterns from previous time slots are concatenated together to form a long pattern vector is used. Then the history is explicitly shown in the input vector. In the third model, an averaging scheme is again used, but with one map to get a first-order representation of the input data. The averaged responses from the first map are used as input patterns for the second map. Thus, the third model consists of a hierarchical structure of maps. It is concluded that the third system is the most interesting because of its accuracy, high tolerance to increasing noise, and high tolerance to the variation of the weighting parameters of the systems

Phonetic typewriter for Finnish and Japanese

A microprocessor-based real-time speech recognition system is described. It is able to produce orthographic transcriptions for arbitrary words or phrases uttered in Finnish or Japanese. It can also be used as a large-vocabulary isolated word recognizer. The acoustic processor of the system transcribing speech into phonemes is based on neural network principles. The so-called phonotopic maps constructed by a self-organizing process are employed. The coarticulation effects in phonetic transcriptions are compensated by means of automatically derived rules which describe the morphology of errors at the acoustic processor output. Without applying any language model, the recognition result is correct up to 92 or even 97 per cent referring to individual letters.<<ETX>>

On-line adaptation in recognition of handwritten alphanumeric characters

We have developed an adaptive online recognizer that is suitable for recognizing isolated alphanumeric characters. It is based on the k nearest neighbor rule. Various dissimilarity measures, all based on dynamic time warping (DTW), have been studied. The main focus of this work is on online adaptation. The adaptation is performed by modifying the prototype set of the classifier according to its recognition performance and the users writing style. These adaptations include: (1) adding new prototypes, (2) inactivating confusing prototypes, and (3) reshaping existing prototypes. The reshaping algorithm is based on learning vector quantization (LVQ). The writers are allowed to use their own natural style of writing, and the adaptation is carried out during normal use in a self-supervised fashion and thus remains otherwise unnoticed by the user.

Phoneme recognition using time-dependent versions of self-organizing maps

Two modifications of the self-organizing map (SOM) are proposed that, unlike the original algorithm, take into account time-dependent features of the input signal. In the first, a time average of a sequence of responses of one SOM is found, and this is recognized by another SOM. In the second, successive input patterns are concatenated together and recognized by the SOM. Comparing the results to those of a recognition system utilizing the original SOM, it was found that one could improve the recognition of isolated phonemes from 10.4% of errors to 7.0% and 5.0% of errors for the integration model and concatenation model, respectively. The improvement in a full-scale system where phoneme segments are also to be located is from 9.2% of errors to 8.2% and 7.6% of errors for the new methods, respectively.<<ETX>>

Self‐organized acoustic feature map in detection of fricative‐vowel coarticulation

The self-organizing map, a neural network algorithm of Kohonen, was used for the detection of coarticulatory variation of fricative [s] preceding vowels [a:], [i:], and [u:]. The results were compared with the psychoacoustic classification of the same samples to find out whether the map had extracted perceptually meaningful features of [s]. The map distinguished samples of [s] in front of [u:] from those in front of [a:] or [i:] throughout the fricative duration. Samples of [s] preceding [a:] and [i:] were distinguished from each other only just before (about 40 ms) the vowel onset. The results agreed with the perceptual classifications. Most judgments (82%) of [s] in front of [u:] were correct, and this variant of [s] was recognized from the first and second halves of segmented fricatives equally well. Samples of [s] in front of [a:] and [i:] were distinguished from each other less accurately. When halves of segmented [s] were perceptually judged, the differentiation between the following [a] and [i] was possible only on the basis of the second half of the fricative. The results demonstrate that the self-organizing map is a useful tool for the extraction of intersubject regularities in speech spectra. The map also provides an easily understandable, on-line, visualization of speech that can be used as feedback in therapy and education.