Extended Self-organizing Map Research Articles

From Seeing to Recognising–An Extended Self-Organizing Map for Human Postures Identification

From Seeing to Recognising–An Extended Self-Organizing Map for Human Postures Identification

An adaptive growing grid model for a non-stationary environment

The self-organizing map (SOM) represents high-dimensional input samples by a 2-dimensional output topological structure, whereby similar input samples are mapped onto the same output unit or neighboring units on a map for visualization. Although many extended SOM models have been proposed, the need to determine a grid structure of SOM before learning, and the lack of adaptability to rapid changes in input data have not yet been fully overcome. This research proposes an adaptive growing grid (AGG) model, which is a novel neural self-organizing map (SOM), for projecting high-dimensional input samples onto an output grid. Due to the need for a grid structure for visualization, the AGG uses both growing and pruning functions and an adaptive learning process in order to adapt its output grid structure and learning function to constantly and rapidly changing input data in a non-stationary environment. The proposed AGG is tested on four basic data sets and one cross-domain data set. In addition, the t-test is used to test whether the proposed AGG outperforms the benchmark model, the growing grid (GG). Based on three evaluation measures, i.e. average quantization error (AQE), topographic error (TE) and dead unit ratio (DUR), the AGG significantly outperforms the GG in a non-stationary environment.

An Extended Self-Organizing Map based on 2-opt algorithm for solving symmetrical Traveling Salesperson Problem

Self-organizing map (SOM) is a powerful variant of neural network for solving optimization problems. Many researchers have reported SOM for Traveling Salesperson Problem; however, problems still exist due to the trapping of the optimization techniques at the local optimal position. In this work, we propose an Extended Self-Organizing Map based on 2-opt algorithm with one-dimensional neighborhood to approach the Symmetrical Traveling Salesperson Problem (STSP). We elaborate our approach for STSP where weights of neurons represent nodes that are placed in the polygonal domain. The selection of winner neuron of SOM has been extended to overcome the problem of trapping of SOM at local optima. The results of SOM are improved through 2-opt local optimization algorithm. We briefly discuss self-organization in neural networks, 2-opt algorithm, and extension applied to SOM. Finally, the algorithm is compared with Kohonen Self-Organizing Map and Evolutionary Algorithm. The results show that our approach performs better as compared to other techniques.

Data driven modeling based on dynamic parsimonious fuzzy neural network

In this paper, a novel fuzzy neural network termed as dynamic parsimonious fuzzy neural network (DPFNN) is proposed. DPFNN is a four layers network, which features coalescence between TSK (Takagi–Sugeno–Kang) fuzzy architecture and multivariate Gaussian kernels as membership functions. The training procedure is characterized by four aspects: (1) DPFNN may evolve fuzzy rules as new training datum arrives, which enables to cope with non-stationary processes. We propose two criteria for rule generation: system error and ε-completeness reflecting both a performance and sample coverage of an existing rule base. (2) Insignificant fuzzy rules observed over time based on their statistical contributions are pruned to truncate the rule base complexity and redundancy. (3) The extended self organizing map (ESOM) theory is employed to dynamically update the centers of the ellipsoidal basis functions in accordance with input training samples. (4) The optimal fuzzy consequent parameters are updated by time localized least square (TLLS) method that exploits a concept of sliding window in order to reduce the computational burden of the least squares (LS) method. The viability of the new method is intensively investigated based on real-world and artificial problems as it is shown that our method not only arguably delivers more compact and parsimonious network structures, but also achieves lower predictive errors than state-of-the-art approaches.

Modeling and Control of a Nonlinear Process Based on the Extended Self-Organizing Map Network

An extended self-organizing map (ESOM) network, which consists of a self-organization phase and an optimization phase, was recently developed to construct a local model network (LMN) automatically using the plant data. However, this previous result suffers two drawbacks: (1) increased computation time in the self-organization phase as the number of local models increases, (2) lack of checking stability conditions for both local models and LMN. To overcome these problems, an improved algorithm for the ESOM network is developed in this paper by employing a competitive learning algorithm for data clustering in the self-organization phase and parametric constraints are formulated in the optimization phase to handle the stability of local models. In addition, the global stability of LMN is addressed. With LMN constructed by the ESOM network, it serves as a basis for building a nonlinear controller that combines several local controllers through the weighting functions obtained by the ESOM algorithm. Literature examples are used to illustrate the proposed ESOM-based modeling and controller design method.

An Extended Self-Organizing Map Network for Modeling and Control of Pulse Jet Fabric Filters

ABSTRACT Pulse jet fabric filters (PJFFs) have become an attractive option of particulate collection utilities, because they can meet stringent particulate emission limits regardless of variation in operating conditions. Despite their wide applications, the present control algorithm for PJFFs can best be described as rudimentary. In this paper, a modeling and control strategy based on the local model network (LMN) is proposed. An extended self-organizing map (ESOM) network is developed to construct the LMN model of the filtration process using the filter's input-output data. Subsequently, these ESOM local models are incorporated into the design of local generalized predictive controllers (GPC), and the proposed controller design is obtained as the weighted sum of these local controllers. Simulation results show that the proposed controller design yields a better performance than both conventional GPC and proportional plus integral (PI) controllers yield.

Modeling of nonlinear systems using extended self-organizing map

Local model networks (LMN) have been shown to be an effective approach for modeling nonlinear systems. Despite recent advances of LMN,a prior knowledge of the process is necessary for determination of LMN structure and weighting function. To overcome these problems, an extended self-organizing map (ESOM) was recently developed. However, the stability of local models is not addressed in this previous work. In this paper, an improved ESOM algorithm is proposed and its application is illustrated by using a CSTR example. Based on the ESOM models obtained, a global PID controller is designed and its robustness is examined. Simulation results verify that ESOM can be used conveniently in modeling and controlling of the processes that are subject to multiple operating regime.

An Extended Self-Organizing Map for Nonlinear System Identification

Local model networks (LMN) are recently proposed for modeling a nonlinear dynamical system with a set of locally valid submodels across the operating space. Despite the recent advances of LMN, a priori knowledge of the processes has to be exploited for the determination of the LMN structure and the weighting functions. However, in most practical cases, a priori knowledge may not be readily accessible for the construction of LMN. In this paper, an extended self-organizing map (ESOM) network, which can overcome the aforementioned difficulties, is developed to construct the LMN. The ESOM is a multilayered network that integrates the basic elements of a traditional self-organizing map and a feed-forward network into a connectionist structure. A two-phase learning algorithm is introduced for constructing the ESOM from the plant input−output data, with which the structure is determined through the self-organizing phase and the model parameters are obtained by the linear least-squares optimization method. Literature examples are used to demonstrate the effectiveness of the proposed scheme.