No full-text available
To read the full-text of this research,
you can request a copy directly from the author.
Neuroclassification of Spatial Data
Abstract
Without any doubt geography is now in the midst of its third quantitative revolution (cf the statistical revolution in the early 1960s, the mathematical modelling revolution in the early 1970s, and now the neurocomputing revolution in the early 1990s). In common with many other areas of science there is a rapidly growing interest in the application of neurocomputing methods. In many ways the driving force is external to the subject in that the tools are being imported rather than developed indigenously. The new tools are also replacements or complements for, or to, existing methods. The general justification is the promise of an improvement in performance and efficiency, fewer critical assumptions, greater ease in handling hard problems, an expansion of the applicability of quantitative and computer based tools and eventually, automation. It is noted that neurocomputing is just one source of new quantitative tools for geography in the 1990s (Openshaw, 1992a).
... Compared with this method, SOMs have the advantage of lower false recognition rates. Moreover, SOMs are suitable for data with a large degree of scattering [17]. ...
The transformer is the most essential component in electrical power transmission. Once a transformer is installed, it is generally used for decades. When failure occurs in a transformer, serious problems can arise, and numerous electrical components that depend on the transformer may be affected. Thus, deterioration diagnosis of transformers is important. The furfural content in the insulating oil of a transformer is widely used as an indicator of transformer deterioration, as furfural is generated by decomposition of the insulating paper. Although the degree of transformer deterioration can be estimated by measuring the furfural content, this measurement is time‐consuming. This paper proposes a novel method for estimating the amount of furfural contained in insulating oil within a short time. The method is realized by combining spectroscopic analysis in the visible region with a pattern recognition technique. The effectiveness of the proposed method is verified through a series of experiments. © 2022 Institute of Electrical Engineers of Japan. Published by Wiley Periodicals LLC.
... Geodemographics research has been developed actively, particularly in Europe and the United States (Harris et al., 2005), and commercialization is progressing. In such studies, general clustering methods, e.g., hierarchical clustering, k-means, and self-organizing maps, which considered robust against outliers (Openshaw, 1994), are frequently applied to spatial data. However, when general clustering methods are applied to spatial data, it is impossible to control the complexity and grouping of regional shapes with homogeneous properties. ...
Cities include districts with various features, and many studies have attempted to grasp the characteristics of cities by clustering districts with similar features. To date, the spatial clustering target has been a single urban space. However, districts with similar characteristics can be observed in different cities. We propose a feature extraction method that generates common feature functions using geographical data, such as population, industry type, use district, and similar districts in two cities. The usefulness of the proposed method was verified by applying the obtained feature function using Tokyo and Kyoto data to another city Osaka as verification data.
... Compared with this method, SOM has the advantage of lower false recognition rates. It is also reported that SOM is suitable for data with large scattering [14]. ...
Induction motor plays an important part in many industrial processes. Detection of motor faults in the early stages improves process productivity, minimizes motor damage, and reduces repair costs. In this paper, a new method is proposed to detect broken rotor bars in a cage induction motor around the rated rotating speed. In this method, first, a frequency analysis is performed according to rotor bar's conditions and the characteristic frequency components are extracted. Further, a two-step diagnosis method based on self-organizing map, which is one of the classification methods, is proposed. Finally, the effectiveness of the proposed diagnosis method is verified by performing some experiments. © 2016 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc. © 2016 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.
... The K-SOM is used to extend current geodemographic practices by the formalisation of spatial relationships between geographical and social space. K-SOM, according to Openshaw (1994a;1995), has been proved more capable for census-data clustering and is superior to k-means. The K-SOM method has also been used by Openshaw and Wymer (1991) to defi ne clusters that rely on socioeconomic, demographic, and health data. ...
In this study we use geographic information systems (GIS) and computational intelligence for geomarketing analysis. GIS technology offers a powerful set of tools for the input, management, and output of data, whereas computational intelligence is used for the analysis and the clustering of data by the use of unsupervised fuzzy clustering and the Gustafson-Kessel algorithm. The advantage of fuzzy geomarketing segmentation is that a customer is not assigned exclusively to one segment only, but rather with a membership value to each cluster. The proposed methodology is applied to the metropolitan area of Athens, Greece. A dataset describes 130 demographic, lifestyle, and economy variables, and the results are analysed and discussed.
... As a well-developed technique, SOM has found many applications in various fields such as data classification, pattern recognition, image analysis, and exploratory data analysis (for an overview, see Oja and Kaski 1999). In the domain of GIS, Openshaw and his colleagues have used the approach in spatial data analysis to carry out the classification of census data (Openshaw 1994, Openshaw et al. 1995. It has been applied to cartographic generalization for building typification (e.g. ...
This paper explores a novel approach to the extraction of spatial objects from the laser-scanning data using an unsupervised clustering technique. The technique, namely self-organizing maps (SOM), creates a set of neurons following a training process based on the input point clouds with attributes of xyz coordinates and the return intensity of laser-scanning data. The set of neurons constitutes a two dimensional planar map, with which each neuron has best match points from an input point cloud with similar properties. Because of its high capacity in data clustering, outlier detection and visualization, SOM provides a powerful technique for the extraction of spatial objects from laser-scanning data. The approach is validated by a case study applied to a point cloud captured using a terrestrial laser-scanning device.
... Self organising maps have previously been applied to geographic datasets. For example to classify vegetation types (Gahegan, 1999) or residential areas (Openshaw, 1994). In these cases the self organising map algorithm was used to clustered samples, simplifying the multivariate complexity of the original dataset, in order to classify geographic regions, represented in conventional cartesian based geographic maps, by vegetative coverage or residential use. ...
Fruit quality and productivity datasets, obtained over two seasons from 24 New Zealand persimmon orchards, were associated with a self organising map trained on the spatial coordinates and geographic region of each orchard. Using this approach, a summary representation of the geographic distribution of the orchards was obtained from the projection plane of the self organising map. By overlaying fruit quality and tree productivity attributes as component planes, spatial patterns between orchards could be observed. In addition, climatic data from regional meteorological stations was associated with the map.
... In the domain of GIS and cartography, relatively few applications have been made. However, Openshaw and his colleagues have used the SOM approach in spatial data analysis to classify census data (Openshaw 1994, Openshaw et al. 1995. Recently, some new proposals have been offered for using SOM to explore spatial data (Li 1998) and in image classification (Luo and Tseng 2000). ...
We propose a novel approach to selection of important streets from a network, based on the technique of a self-organizing map (SOM), an artificial neural network algorithm for data clustering and visualization. Using the SOM training process, the approach derives a set of neurons by considering multiple attributes including topological, geometric and semantic properties of streets. The set of neurons constitutes a SOM, with which each neuron corresponds to a set of streets with similar properties. Our approach creates an exploratory linkage between the SOM and a street network, thus providing a visual tool to cluster streets interactively. The approach is validated with a case study applied to the street network in Munich, Germany.
... see Bacao, Lobo, & Painho (2008), Kohonen (1997), Kohonen, Hynninen, Kangas, and Laaksonen (1996), Openshaw (1989) ...
There is a long cartographic tradition of describing cities through a focus on the characteristics of their residents. A review of the history of this type of urban social analysis highlights some persistent challenges. In this paper existing geodemographic approaches are extended through coupling the Kohonen Self-Organizing Map algorithm (SOM), a data-mining technique, with geographic information systems (GIS). This approach allows the construction of linked maps of social (attribute) and geographic space. This novel type of geodemographic classification allows ad hoc hierarchical groupings and exploration of the relationship between social similarity and geographic proximity. It allows one to filter complex demographic datasets and is capable of highlighting general social patterns while retaining the fundamental social fingerprints of a city. A dataset describing 79 attributes of the 2217 census tracts in New York City is analyzed to illustrate the technique. Pairs of social and geographic maps are formally compared using simple pattern metrics. Our analysis of New York City calls into question some assumptions about the functional form of spatial relationships that underlie many modeling and statistical techniques.
... In the domain of GIS and cartography, relatively few applications have been made. However, Openshaw and his colleagues have used the SOM approach in spatial data analysis to classify census data (Openshaw 1994, Openshaw et al. 1995. Recently, some new proposals have been offered for using SOM to explore spatial data (Li 1998) and in image classification (Luo and Tseng 2000). ...
We propose a novel approach to selection of important streets from a network, based on the technique of a self-organizing map (SOM), an artificial neural network algorithm for data clustering and visualization. Using the SOM training process, the approach derives a set of neurons by considering multiple attributes including topological, geometric and semantic properties of streets. The set of neurons constitutes a SOM, with which each neuron corresponds to a set of streets with similar properties. Our approach creates an exploratory linkage between the SOM and a s treet network, thus providing a visual tool to cluster streets interactively. The approach is validated with a case study applied to the street network in Munich, Germany.
... A recent book edited by Agarwal and Skupin (2007) is an excellent compilation of various applications of SOMs in GIS. As argued by Openshaw (1989, page 59 ...
Given that many spatial interaction (SI) systems are often constituted in large databases with high thematic dimensionality, data complexity reduction tasks are essential. The opportunity exists for researchers to examine the formation of different types of SIs as well as their interdependencies by exploring the patterns embedded in the data. To circumvent the limitations of existing methods of flow data compression and visual exploration, we propose an integrated computational and visual approach, known as VISIDAMIN, for handling both SI data projection and SI data quantization at once. The computational method of self-organizing maps serves as the data mining engine in this process. Using a large domestic air travel dataset as a case study, we examine how the characteristics of the air transport system interact with the SI system to create relationships and structures within the US domestic airline market.
We have examined the noncontact respiration measurement by applying the active stereo for realizing easy screening of chronic obstructive pulmonary disease (COPD). In this study, we apply the self-organizing map (SOM) for classification of unforced respiratory waveform. The automatic classification of the respiratory waveform is based on the difference in the undulating pattern of respiratory waveform between the COPD patients and the healthy subjects. As the result of the experiment, it became clear that the classification by SOM is almost equivalent to the classification by respiratory specialists.
Artificial neural networks (ANNs) have attracted the attention of researchers in many fields, and have been used to solve a wide range of problems. In the field of remote sensing they have been used in a variety of applications, including land cover mapping, image compression, geological mapping and meteorological image classification, and have generally proved to be more powerful than conventional statistical classifiers, especially when training data are limited and the data in each class are not normally distributed.
The use of ANNs requires some critical decisions on the part of the user. These decisions, which are mainly concerned with the determinations of the components of the network structure and the parameters defined for the learning algorithm, can significantly affect the accuracy of the resulting classification. Although there are some discussions in the literature regarding the issues that affect network performance, there is no standard method or approach that is universally accepted to determine the optimum values of these parameters for a particular problem.
In this thesis, a feed-forward network structure that learns the characteristics of the training data through the backpropagation learning algorithm is employed to classify land cover features using multispectral, multitemporal, and multisensory image data. The thesis starts with a review and discussion of general principles of classification and the use of artificial neural networks. Special emphasis is put on the issue of feature selection, due to the availability of hyperspectral image data from recent sensors. The primary aims of this research are to comprehensively investigate the impact of the choice of network architecture and initial parameter estimates, and to compare a number of heuristics developed by researchers. The most effective heuristics are identified on the basis of a large number of experiments employing two real-world datasets, and the superiority of the optimum settings using the 'best' heuristics is then validated using an independent dataset. The results are found to be promising in terms of ease of design and use of ANNs, and in producing considerably higher classification accuracies than either the maximum likelihood or neural network classifiers constructed using ad hoc design and implementation strategies. A number of conclusions are drawn and later used to generate a comprehensive set of guidelines that will facilitate the process of design and use of artificial neural networks in remote sensing image classification.
This study also explores the use of visualisation techniques in understanding the behaviour of artificial neural networks and the results produced by them. A number of visual analysis techniques are employed to examine the internal characteristics of the training data. For this purpose, a toolkit allowing the analyst to perform a variety of visualisation and analysis procedures was created using the MATLAB software package, and is available in the accompanying CD-ROM. This package was developed during the course of this research, and contains the tools used during the investigations reported in this thesis.
The contribution to knowledge of the research work reported in this thesis lies in the identification of optimal strategies for the use of ANNs in land cover classifications based on remotely sensed data. Further contributions include an indepth analysis of feature selection methods for use with high-dimensional datasets, and the production of a MATLAB toolkit that implements the methods used in this study.
We have examined the non-contact respiration measurement by applying the active stereo for realizing easy screening of Chronic Obstructive Pulmonary Disease (COPD). In this study, we apply the Self-Organizing Map (SOM) for classification of unforced respiratory waveform. The automatic classification of the respiratory waveform is based on the difference in the undulating pattern of respiratory waveform between the COPD patients and the healthy subjects. As the result of the experiment, it becomes clear that the classification by SOM is almost equivalent to the classification by respiratory specialists.
The self-organizing map (SOM) method has become popular in various disciplines for visual exploration of large, complex, multivariate databases. Despite the increasing popularity of this powerful, but computationally intensive, neural network approach, web applications involving SOM construction and visualization are still rare. We introduce SOMViz, a simple, interactive, web-based prototype system, accessible online through a simple graphical user interface, allowing users without programming knowledge to generate SOMs from multivariate geographic datasets, and visually explore them in combination with thematic maps. At the heart of our contribution lies a proof-of-concept based on an open architecture, including open source technology. We first outline and evaluate design decisions of the proof-of-concept system already completed in 2009 in the context of current state-of-the-art web technology, and then present implementation details and respective functionality of the SOMViz prototype applied to a case study with Swiss population census data as an example. With this contribution, we hope on the one hand to facilitate access to the powerful SOM analysis method for non-specialists, based on generic and open web technology. On the other hand, we offer this solution as a starting point to apply long-standing cartographic design principles and approaches to emerging non-geographic mapping technologies and methods.
Correct description of fuel properties is critical to improve fire danger assessment and fire behaviour modeling, since they guide both fire ignition and fire propagation. This chapter deals with properties of fuel that can be considered static in short periods of time: biomass loads, plant geometry, compactness, etc. Mapping these properties require a detail knowledge of vegetation vertical and horizontal structure. Several systems to classify the great diversity of vegetation characteristics in few fuel types are described, as well as methods for mapping them with special emphasis on those based on remote sensing images.
There are two important problems in clustering of small area statistics: handling of outliers in small area statistics and complexity of spatial distribution of typologies created by classifying small areas. The purpose of this paper is to show a procedure of geographical clustering using small area statistics in order to solve these problems. According to the results of two examinations, Self-Organizing Maps (SOM) constraining the range of updating weights is better classifier than K-means. As to the issue of simplifying spatial distribution of typologies, we showed that there are relations between the level of the spatial smoothing and the spatial extent of the study area.
The climate of a particular region is governed by factors that may be remote, such as the El Nino Southern Oscillation or local, such as topography. However, the daily weather characteristics of a region are controlled by the synoptic-scale atmospheric state. Therefore changes in the type, frequency, duration or intensity of particular synoptic states over a region would result in changes to the local weather and long-term climatology of the region. The relationship between synoptic-scale circulation and the rainfall response is examined for a 31-year period at two stations in different rainfall regimes in South Africa. Dominant rain-bearing synoptic circulations are identified for austral winter and summer as mid-latitude cyclones and convective systems respectively whereas no circulations are dominantly associated with spring and autumn rainfall. Over the 31-year period a statistically significant increase in the frequency of characteristic summer circulation modes is observed during summer, winter and spring. During autumn a statistically significant shift towards characteristically winter circulation modes is evident. Seasonal rainfall trends computed at each station corroborate those of the circulation data. Extreme rainfall is associated with particular circulation modes and trends in both circulation and station data show an earlier occurrence of extreme rainfall during the rainy season.
This paper endeavours to put the discussion on errors and uncertainties in geographical information systems (GISs) in a more
systematic way by examining the strength and weakness of discrete objects and continous fields, the two distinct schools of
spatial data modelling. In doing so, it argues that neither discrete objects nor continous fields alone provide objective
and complete representations of highly complex geographical phenomena, though there are good reasons for asserting that continuous
fields are better suited to modelling spatial dependence, heterogeneity and fuzzines significant in geographical reality than
discrete objects. Thus, there seems to be merit in adopting an integrated model incorporating analytical capabilities of fields
and generalization functions of objects, for which extended TIN (triangulated irregular network) models along with their duals
(Voronoi diagrams) provide a pragmatical solution.
The paper describes the development of Kohonen-net-based methods suitable for the classification of large spatial datasets suitable for parallel processing. Parallelising the Kohonen net is not easy because the degree of natural parallelism is finely grained. This paper presents a new algorithm and demonstrates its performance on the Cray T3D parallel supercomputer.
Artificial neural networks (ANNs) have become a popular tool in the analysis of remotely sensed data. Although significant progress has been made in image classification based upon neural networks, a number of issues remain to be resolved. This paper reviews remotely sensed data analysis with neural networks. First, we present an overview of the main concepts underlying ANNs, including the main architectures and learning algorithms. Then, the main tasks that involve ANNs in remote sensing are described. The limitations and crucial issues relating to the application of the neural network approach are discussed. A brief review of the implementation of ANNs in some of the most popular image processing software packages is presented. Finally, we discuss the application perspectives of neural networks in remote sensing image analysis.
This study aims to introduce contextual Neural Gas (CNG), a variant of the Neural Gas algorithm, which explicitly accounts for spatial dependencies within spatial data. The main idea of the CNG is to map spatially close observations to neurons, which are close with respect to their rank distance. Thus, spatial dependency is incorporated independently from the attribute values of the data. To discuss and compare the performance of the CNG and GeoSOM, this study draws from a series of experiments, which are based on two artificial and one real-world dataset. The experimental results of the artificial datasets show that the CNG produces more homogenous clusters, a better ratio of positional accuracy, and a lower quantization error than the GeoSOM. The results of the real-world dataset illustrate that the resulting patterns of the CNG are theoretically more sound and coherent than that of the GeoSOM, which emphasizes its applicability for geographic analysis tasks.
Teaching methodological courses means providing students with general ideas to think with, rather than specific facts to think about. Rethinking the teaching task must be informed by an awareness of ideological, philosophical, technical and analytical perspectives, and should look for basic themes uniting apparently different approaches. Almost all quantitative methodologies are actually specific examples of the mathematical operation of mapping. Genuine understanding is achieved by visualisation of a problem, rather than algebraic manipulation, and the teaching task could be aided greatly by employing the power of interactive computer graphics. It is suggested that the analytical and pedagogic approaches used by experienced teachers might be shared initially in an international conference, and so form the basis for more effective and efficient teaching in the future.
This paper presents a methodology for the creation of homogeneous demographic regions with geographical information systems (GIS) and computational intelligence. The proposed method is unsupervised fuzzy classification performed by neural networks using the fuzzy Kohonen algorithm. GIS technology offers a powerful set of tools for the input, management, and output of data, whereas computational intelligence is used for the analysis and the classification of the data. The proposed methodology is applied to the municipality of Athens, in Greece. Finally the advantages and disadvantages of the approach are discussed.
To describe how geodemographic segmentation systems might be useful as a quick and easy way of exploring postcoded health databases for potential interesting patterns related to deprivation and other socioeconomic characteristics.
This is demonstrated using GB Profiles, a freely available geodemographic classification system developed at Leeds University. It is used here to screen a database of colorectal cancer registrations as a first step in the analysis of that data.
Conventional geodemographics is a fairly simple technology and a number of outstanding methodological problems are identified. A solution to some problems is illustrated by using neural net based classifiers and then by reference to a more sophisticated geodemographic approach via a data optimal segmentation technique.
Neural networks are introduced which can be taught by classical or instrumental conditioning to fire in response to arbitrary learned classes of patterns. The filters of output cells are biased by presetting cells whose activation prepares the output cell to expect prescribed patterns. For example, an animal that learns to expect food in response to a lever press becomes frustrated if food does not follow the lever press. It's expectations are thereby modified, since frustration is negatively reinforcing. A neural analog with aspects of cerebellar circuitry is noted, including diffuse mossy fiber inputs feeding parallel fibers that end in Purkinje cell dendrites, climbing fiber inputs ending in Purkinje cell dendrites and giving off collaterals to nuclear cells, and inhibitory Purkinje cell outputs to nuclear cells. The networks are motivated by studying mechanisms of pattern discrimination that require no learning. The latter often use two successive layers of inhibition, analogous to horizontal and amacrine cell layers in vertebrate retinas. Cells exhibiting hue (in)constancy, brightness (in)constancy, or movement detection properties are included. These results are relevant to Land's retinex theory and to the existence of opponent- and nonopponent-type cell responses in retinal cells. Some adaptation mechanisms, and arousal mechanisms for crispening the pattern weights that can fire a given cell, are noted.
A nerve net model for the visual cortex of higher vertebrates is presented. A simple learning procedure is shown to be sufficient for the organization of some essential functional properties of single units. The rather special assumptions usually made in the literature regarding preorganization of the visual cortex are thereby avoided. The model consists of 338 neurones forming a sheet analogous to the cortex. The neurones are connected randomly to a retina of 19 cells. Nine different stimuli in the form of light bars were applied. The afferent connections were modified according to a mechanism of synaptic training. After twenty presentations of all the stimuli individual cortical neurones became sensitive to only one orientation. Neurones with the same or similar orientation sensitivity tended to appear in clusters, which are analogous to cortical columns. The system was shown to be insensitive to a background of disturbing input excitations during learning. After learning it was able to repair small defects introduced into the wiring and was relatively insensitive to stimuli not used during training.
Adaptive resonance architectures are neural networks that self-organize stable pattern recognition codes in real-time in response to arbitrary sequences of input patterns. This article introduces ART 2, a class of adaptive resonance architectures which rapidly self-organize pattern recognition categories in response to arbitrary sequences of either analog or binary input patterns. In order to cope with arbitrary sequences of analog input patterns, ART 2 architectures embody solutions to a number of design principles, such as the stability-plasticity tradeoff, the search-direct access tradeoff, and the match-reset tradeoff. In these architectures, top-down learned expectation and matching mechanisms are critical in self-stabilizing the code learning process. A parallel search scheme updates itself adaptively as the learning process unfolds, and realizes a form of real-time hypothesis discovery, testing, learning, and recognition. After learning self-stabilizes, the search process is automatically disengaged. Thereafter input patterns directly access their recognition codes without any search. Thus recognition time for familiar inputs does not increase with the complexity of the learned code. A novel input pattern can directly access a category if it shares invariant properties with the set of familiar exemplars of that category. A parameter called the attentional vigilance parameter determines how fine the categories will be. If vigilance increases (decreases) due to environmental feedback, then the system automatically searches for and learns finer (coarser) recognition categories. Gain control parameters enable the architecture to suppress noise up to a prescribed level. The architecture's global design enables it to learn effectively despite the high degree of nonlinearity of such mechanisms.
A model for a large network of "neurons" with a graded response (or sigmoid input-output relation) is studied. This deterministic system has collective properties in very close correspondence with the earlier stochastic model based on McCulloch - Pitts neurons. The content- addressable memory and other emergent collective properties of the original model also are present in the graded response model. The idea that such collective properties are used in biological systems is given added credence by the continued presence of such properties for more nearly biological "neurons." Collective analog electrical circuits of the kind described will certainly function. The collective states of the two models have a simple correspondence. The original model will continue to be useful for simulations, because its connection to graded response systems is established. Equations that include the effect of action potentials in the graded response system are also developed.
Computational properties of use of biological organisms or to the construction of computers can emerge as collective properties of systems having a large number of simple equivalent components (or neurons). The physical meaning of content-addressable memory is described by an appropriate phase space flow of the state of a system. A model of such a system is given, based on aspects of neurobiology but readily adapted to integrated circuits. The collective properties of this model produce a content-addressable memory which correctly yields an entire memory from any subpart of sufficient size. The algorithm for the time evolution of the state of the system is based on asynchronous parallel processing. Additional emergent collective properties include some capacity for generalization, familiarity recognition, categorization, error correction, and time sequence retention. The collective properties are only weakly sensitive to details of the modeling or the failure of individual devices.
The authors present a neural network architecture, called ARTMAP,
that autonomously learns to classify arbitrarily many, arbitrarily
ordered vectors into recognition categories based on predictive success.
This supervised learning system is built up from a pair of adaptive
resonance theory modules that are capable of self-organizing stable
recognition categories in response to arbitrary sequences of input
patterns. Tested on a benchmark machine learning database in both online
and offline simulations, the ARTMAP system learned orders of magnitude
more quickly, efficiently, and accurately than alternative algorithms,
and achieves 100% accuracy after training on less than half the input
patterns in the database. It achieves these properties by using an
internal controller that conjointly maximizes predictive generalization
and minimizes predictive error by linking predictive success to category
size on a trial-by-trial basis, using only local operations
Neurocomputing has the potential to revolutionise many areas of urban and regional modelling by providing a general purpose systems modelling tool in applications where data exist. This chapter examines the empirical performance of a feedforward neural net as the basis for representing the spatial interaction contained within journey to work data. The performance of the neural net representation is compared with various types of conventional model. It is concluded that there is considerable potential for many more neural net applications in this and related areas.
An abstract is not available.
An abstract is not available.
Packed with real-time computer simulations and rigorous demonstrations of these phenomena, this book includes results on vision, speech, cognitive information processing, adaptive pattern recognition, adaptive robotics, conditioning and attention, cognitive-emotional interactions, and decision making under risk.
"Neural Networks and Natural Intelligence" first discusses neural network architecture for preattentive 3-D vision and then shows how this architecture provides a unified explanation, through systematic computer simulations, of many classical and recent phenomena from psycho-physics, visual perception, and cortical neurophysiology. It illustrates within the domain of preattentive boundary segmentation and featural filling-in, how computer experiments help to develop and refine computational vision models. (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Este libro ofrece un nuevo enfoque a la auto-organización, adapta -ción, aprendizaje y memoria, que da lugar a cursos para postgraduados en ciencias de la información, ciencias computacionales, psicología, biología teórica, y física.
Described here is sparse distributed memory (SDM) as a neural-net associative memory. It is characterized by two weight matrices and by a large internal dimension - the number of hidden units is much larger than the number of input or output units. The first matrix, A, is fixed and possibly random, and the second matrix, C, is modifiable. The SDM is compared and contrasted to (1) computer memory, (2) correlation-matrix memory, (3) feet-forward artificial neural network, (4) cortex of the cerebellum, (5) Marr and Albus models of the cerebellum, and (6) Albus' cerebellar model arithmetic computer (CMAC). Several variations of the basic SDM design are discussed: the selected-coordinate and hyperplane designs of Jaeckel, the pseudorandom associative neural memory of Hassoun, and SDM with real-valued input variables by Prager and Fallside. SDM research conducted mainly at the Research Institute for Advanced Computer Science (RIACS) in 1986-1991 is highlighted.
The last 5 years have seen the development of artificial intelligence (AI) methods that are capable of being applied to many practical problems. This paper looks at some of the actual and potential applications for AI in the area of space analysis and modeling relevant to GIS. Of particular importance here is the development of data exploration tools for pattern description, relationship seeking, and modeling. The paper reviews some of the applicable heuristic search, artificial life, genetic optimization, and neurocomputing methods.
Theoretical models of the human brain and proposed neural-network computers are developed analytically. Chapters are devoted to the mathematical foundations, background material from computer science, the theory of idealized neurons, neurons as address decoders, and the search of memory for the best match. Consideration is given to sparse memory, distributed storage, the storage and retrieval of sequences, the construction of distributed memory, and the organization of an autonomous learning system.
This paper analyses a model for the parallel development and adult coding of neural feature detectors. The model was introduced in Grossberg (1976). We show how experience can retune feature detectors to respond to a prescribed convex set of spatial patterns. In particular, the detectors automatically respond to average features chosen from the set even if the average features have never been experienced. Using this procedure, any set of arbitrary spatial patterns can be recoded, or transformed, into any other spatial patterns (universal recoding), if there are sufficiently many cells in the network's cortex. The network is built from short term memory (STM) and long term memory (LTM) mechanisms, including mechanisms of adaptation, filtering, contrast enhancement, tuning, and nonspecific arousal. These mechanisms capture some experimental properties of plasticity in the kitten visual cortex. The model also suggests a classification of adult feature detector properties in terms of a small number of functional principles. In particular, experiments on retinal dynamics, including amarcrine cell function, are suggested.
Stability and encoding properties of two-layer nonlinear feedback
neural networks are examined. Bidirectionality is introduced in neural
nets to produce two-way associative search for stored associations. The
bidirectional associative memory (BAM) is the minimal two-layer
nonlinear feedback network. The author proves that every n -by-
p matrix M is a bidirectionally stable
heteroassociative content-addressable memory for both binary/bipolar and
continuous neurons. When the BAM neutrons are activated, the network
quickly evolves to a stable state of two-pattern reverberation, or
resonance. The stable reverberation corresponds to a system energy local
minimum. Heteroassociative information is encoded in a BAM by summing
correlation matrices. The BAM storage capacity for reliable recall is
roughly m <min ( n , p ). It is also shown that
it is better on average to use bipolar {-1,1} coding than binary {0.1}
coding of heteroassociative pairs. BAM encoding and decoding are
combined in the adaptive BAM, which extends global bidirectional
stability to real-time unsupervised learning
Various information-processing capabilities of self-organizing nets of threshold elements are studied. A self-organizing net, learning from patterns or pattern sequences given from outside as stimuli, "remembers" some of them as stable equilibrium states or state-transition sequences of the net. A condition where many patterns and pattern sequences are remembered in a net at the same time is shown. The stability degree of their remembrance and recalling under noise disturbances is investigated theoretically. For this purpose, the stability of state transition in an autonomous logical net of threshold elements is studied by the use of characteristics of threshold elements.
Invariant Pattern Recognition and Recall by an Attentive Self-Organizing ART Architecture in a Non-Stationary World
- G Carpenter
- S Grossberg
A Regionalization Procedure for a Comparative Regional Taxonomy of the UK
- S Openshaw
A Comparative Evaluation of Three Neuroclassifiers of Census Data
- S Openshaw
- C Wymer
- M E Charlton
Feature Discovery by Competitive Learning
- D Rumelhart
- D Zipster
Further Thoughts on Geography and GIS: A Reply
- S Openshaw
Special Classifications An Introductory Guide to the 1991 Census
- S Openshaw
Multivariate Analysis of Census Data: The Classification of Areas
- S Openshaw
A Review of the Opportunities and Problems in Applying Neurocomputing Methods to Marketing Applications
- S Openshaw
A Self-Organizing ARTMAP Neural Architecture for Supervised Learning and Pattern Recognition
- G A Carpenter
- S Grossberg
- J H Reynolds