A modified training scheme for SOFM to cluster multispectral images

We applied Kohonen's self-organizing feature maps (SOM) into the problem of selecting the buttons. 3D data spaces are represented by two maps of 2D SOM which are formed on the planes that are projected from 3D objects. Assuming the ability of SOM to approximate the shape of input-data spaces, we select the best matching color-series when reference-data point exists in the SOM. Using this method, very good recognition (96%) was obtained. This method is better than the neural network with backpropagation learning on the points that adjustments of parameters are easy and paradoxical data-relations are permitted.

Spiking Self-organizing Maps for Classification Problem

We examine the performance of several neural network vector quantization (VQ) methods on image coding. The VQ methods we look at are Kohonen's Self-Organizing Feature Map (KSOFM), Frequency-Sensitive Competitive Learning (FSCL), and Self-Creating and Organizing Neural Network (SCONN). We also look at variations of these algorithms by combining different methods. Our simulation results show that the best performance is achieved by the SCONN and the combination KSOFM and FSCL.

The paper presents the design of municipal creditworthiness parameters. Further, a model is designed based on Learning Vector Quantization neural networks for municipal creditworthiness classification. The model is composed of Kohonen's Self-organizing Feature Maps (unsupervised learning) whose outputs represent the input of the Learning Vector Quantization neural networks (supervised learning).

In the traditional self-organizing map (SOM) the best matching unit (BMU) affects other neurons, through the learning rule, as a function of distance. Here, we propose a new parameter in the learning rule so neurons are not only affected by BMU as a function of distance, but as a function of the frequency of activation from both, the BMU and input vectors, to the affected neurons. This frequency parameter allows non radial neighborhoods and the quality of the formed maps is improved with respect to those formed by traditional SOM, as we show by comparing several error measures and five data sets.KeywordsInput VectorLearning RuleError QuantizationFrequency FunctionLearning FactorThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

An investigation of local energy surface detection integrated with neural network techniques for image segmentation is presented, as applied in the feature extraction of chromosomes from image datasets obtained using an experimental confocal microscope. Use of the confocal microscope enables biologists to observe dividing cells (living or preserved) within a three-dimensional (3-D) volume, that can be visualised from multiple aspects, allowing for increased structural insight. The Nomarski differential interference contrast mode used for imaging translucent specimens, such as chromosomes, produces images not suitable for volume rendering. Segmentation of the chromosomes from this data is, thus, necessary. A neural network based on competitive learning, known as Kohonen's self-organizing feature map (SOFM) was used to perform segmentation, using a collection of statistics or features defining the image. Our past investigation showed that standard features such as the localized mean and variance of pixel intensities provided reasonable extraction of objects such as mitotic chromosomes, but surface detail was only moderately resolved. In this current work, a biologically inspired feature known as local energy is investigated as an alternative image statistic based on phase congruency in the image. This, along with different combinations of other image statistics, is applied in a SOFM, producing 3-D images exhibiting vast improvement in the level of detail and clearly isolating the chromosomes from the background. Index Terms-DIC, differential interference contrast, feature extraction, feature space, image segmentation, local energy, Morlet wavelet, phase congruency, self organizing feature map, SOFM.

Self‐organizing map (SOM) has been studied as a model of map formation in the brain cortex. However, the original model present several oversimplifications. For example, neurons in the cortex present a refractory period in which they are not able to be activated, restriction that should be included in the SOM if a better model is to be achieved. Although several modifications have been studied in order to include this biological restriction to the SOM, they do not reflect biological plausibility. Here, we present a modification in the SOM that allows neurons to enter a refractory period (SOM‐RP) if they are the best matching unit (BMU) or if they belong to its neighborhood. This refractory period is the same for all affected neurons, which contrasts with previous models. By including this biological restriction, SOM dynamics resembles in more detail the behavior shown by the cortex, such as non‐radial activity patterns and long distance influence, besides the refractory period. As a side effect, two error measures are lower in maps formed by SOM‐RP than in those formed by SOM.

Segmentation and grading of brain tumors on apparent diffusion coefficient images using self-organizing maps

Summary Well-test data have been used traditionally for determining a variety of reservoir parameters, such as average permeability, storage capacity, reservoir damage, presence of faults and fractures, and reservoir mechanism. A number of techniques, both conventional methods, such as type-curve matching and numerical simulation, and artificial intelligence (AI) methods, have been used for identifying well-test models. These methods are laborious and time-consuming and at times give incorrect results. Artificial neural networks (ANN's) are recent developments in computer vision and image analysis. These networks are specialized computer software that generate a strategy to produce nonlinear mapping functions for complex problems. ANN's are commonly used as a tool for recognizing an object or predicting an event given an associated pattern. Only a limited number of applications of ANN for analyzing well-test data have been reported. These applications are mostly model-specific (developed for specific reservoir models) and, hence, are not general enough. This paper presents a new method based on ANN's that uses Ko-honen's self-organizing feature (SOF) mapping technique to identify well-test interpretation models. By grouping well-test data into distinct categories, the SOF algorithm produces a general mapping function. This method can help analyze well-test data from a large variety of reservoirs (including reservoirs with faults, fractures, boundaries, etc.) more efficiently and inexpensively than was feasible previously.

This paper addresses the problem of visual assessment of clustering tendency in p-dimensional data using two extensions of Kohonen's self-organizing feature map (SOFM). We show that SOFM cell displays generally do not produce visual evidence that leads to good guesses about cluster substructure or data density even for 2-dimensional data. The two proposed extensions of SOFM improve the quality of displays and enable us to make better guesses about the existence of substructure in data.

Proposes a new scheme for designing a nearest-prototype classifier. The system starts with the minimum number of prototypes, equal to the number of classes. Kohonen's self-organizing feature map (SOFM) algorithm is used to obtain this initial set of prototypes. Then, on the basis of the classification performance, new prototypes are generated dynamically, similar prototypes are merged, and prototypes with less significance are deleted, leading to better performance. If prototypes are deleted or new prototypes appear, then they are retrained using Kohonen's SOFM algorithm with the winner-only update scheme. This adaptation continues until the system satisfies a termination condition. The classifier has been tested with several well-known data sets. The results obtained are quite satisfactory.

In Kohonen’s Self-Organizing Maps (SOM) algorithm, preserving the map structure to represent the real input patterns appears to be a significant process. Misinterpretation of the training samples can lead to failure in identifying the important features that may affect the outcomes generated by the SOM model. This paper presents detail explanation on SOM learning algorithm and its applications. Some issues related to SOM’s architecture are also discussed, namely the formulation of training data from input samples, and the Best Matching Unit (BMU) identification for better visualization of large datasets, and improvement made to the SOM algorithm.

Research on artificial neural networks (ANNs) has been carried out for more than five decades. A renewed interest appeared in the 80's with the finding of powerful models like J. Hopfield's recurrent networks, T. Kohonen's self-organizing feature maps, and the back-propagation rule. At that time, there was no platform that was at the same time versatile enough for any ANN model to be implemented and fast enough to solve large problems. Super-computers were the sole exception to this rule, but were prohibitively expensive for most applications. However, both research scientists and application engineers clearly identified the need for such a computing power. This triggered many projects in the field. In parallel, research on multi-processor systems started during the 60's. Systolic arrays have been proposed in 1979 as a means to fully exploit the possibilities of VLSI. Two previous theses, by F. Blayo and C. Lehmann, have studied the use of bi-dimensional systolic arrays for neural computation. At first, the presented system, called GENES, has been designed for the Hopfield model. Extensions to other ANNs have also been proposed. The goal of the present thesis is to study, design, build, and analyze an efficient accelerator for neural computation. In a first step, the GENES architecture has been extended towards generality and efficiency. This includes a thorough analysis of ANN models, of other neural computers, and of previous GENES implementations. The result of this work is the GENES IV integrated circuit, whose architecture has been co-designed by P. Ienne and the author. The main part of this thesis discusses the architecture, the design and the analysis of the MANTRA I machine, a neural computer based on a GENES IV array with up to 40 × 40 processing elements (PEs). The delta rule (and hence the Perceptron and Adaline rules), the back-propagation rule, the Hopfield model, and the Kohonen model can be implemented on this system. Although not a generic system, such a machine may be regarded as a multi-model neural computer. A prototype has been running for a year and is used daily by software designers. Several novel features distinguish the MANTRA I machine from other neural systems. First, it belongs to the few existing neural computers, contrary to the majority of implementations, which are specific to an application or to an algorithm. The machine does not hard-wire any algorithm, but provides the necessary primitives to implement the target models. This is a key feature for research, since several algorithms or versions of an algorithm can be tested on a problem. It is an important aspect for applications as well, because different ANN models are often cascaded to solve a problem. The GENES IV array — that is, the computing core of the MANTRA I machine — features synapse-level parallelism (i.e., one real or virtual PE is allocated per synapse or neural connection), while most other systems exploit only neuron-level parallelism (i.e., one PE per neuron). Hence, this system aims at a much finer parallelism grain and is well suited for massively parallel architectures. The problem size that can be computed by a neural accelerator should not be limited by the hardware (except for memory size). Therefore, it is essential to support time-sharing of PEs. On the MANTRA I machine, this is achieved by the concept of virtual arrays. Matrices are divided into sub-matrices that can be mapped onto the physical array, which is then time-shared among them. An efficient mechanism has been implemented to swap sub-matrices in background, while some other computation is performed. Since systolic arrays are pipelined systems, it is important to avoid emptying and re-filling them too often, in order to keep the hardware utilization rate high. Therefore, a systolic instruction flow has been implemented, so that each instruction follows the data for which it has been issued. Like any SIMD system, the MANTRA I machine is composed of a parallel or SIMD module and a control module. The SIMD module consists of the GENES IV array and a set of dedicated units designed for the computation that scales with the number of neurons and would poorly fit on a bi-dimensional array. A complex system of FIFOs and memories sustains the required input/output streams for the systolic array. The control module is a complete SISD system. Its tasks are (1) to control the SIMD module by dispatching instructions, (2) to manage data input and output, (3) to communicate with the external world, and (4) to perform data pre- and post-processing. The SIMD instructions are of the very long instruction word (VLIW) type. Synchronization between the two modules is achieved by an instruction FIFO. The performance of the MANTRA I machine has been analyzed using the delta rule. Measurements show that the sustained performance is very close to its peak value, as long as the problem fits in the memory banks connected to the GENES IV array. Experiments have also been run to investigate the impact of the constraints imposed by the hardware on the convergence of algorithms. Finally, the use of systolic arrays as neural accelerators is discussed in the light of the experience acquired with the GENES IV array and the MANTRA I machine. The weaknesses of the machine are analyzed, and several solutions are proposed to avoid them in a future design. A general discussion of the future of neural computers concludes this thesis.

One of the most prominent Neural Networks (NNs) reported in the literature is the Kohonen’s Self-Organizing Map (SOM). In spite of all its desirable capabilities and the scores of reported applications, it, unfortunately, possesses some fundamental drawbacks. Two of these handicaps are the quality of the map learned and the time required to train it. The most demanding phase of the algorithm involves determining the so-called Best Matching Unit (BMU), which requires time that is proportional to the number of neurons in the NN. The focus of this paper is to reduce the time needed for this tedious task, and to attempt to obtain an approximation of the BMU is as little as logarithmic time. To achieve this, we depend heavily on the work of [3,6], where the authors focused on how to accurately learn the data distribution connecting the neurons on a self-organizing tree, and how the learning algorithm, called the Tree-based Topology-Oriented SOM (TTOSOM), can be useful for data clustering [3,6] and classification [5]. We briefly state how we intend to reduce the training time for identifying the BMU efficiently. First, we show how a novel hyperplane-based partitioning scheme can be used to accelerate the task. Unlike the existing hyperplane-based partitioning methods reported in the literature, our algorithm can avoid ill-conditioned scenarios. It is also capable of considering data points that are dynamic. We demonstrate how these hyperplanes can be recursively defined, represented and computed, so as to recursively divide the hyper-space into two halves. As far as we know, the use of random hyperplanes to identify the BMU is both pioneering and novel.

The noise label of data is a problem that can cause low performance of deep learning. It is difficult to manually relabel due to huge amounts of data. In addition, there are much more problems due to the similarity of sounds that are difficult to manually distinguish and label sound data. We proposed a data cleaning method using SOM (Self-Organizing Map), one of the unsupervised learning methods. In order to extract compact features from audio, densely connected layer with log scaled Mel-spectrogram is used. Data selection is performed based on the Euclidean distance of each Best matching unit (BMU) derived through the SOM. We also experiment with various grid sizes for SOM to find an efficient grid size. In addition, an appropriate distance finding experiment is conducted. This method is evaluated in sound classification using a pre-trained DenseNet model.