Wyniki wyszukiwania - BazTech

1

Convergence Analysis of Inverse Iterative Neural Networks with L₂ Penalty

Wen Y., Wang J., Huang B., Zurada J. M.

Journal of Applied Computer Science Methods

|

2016

|

Vol. 8 No. 2

85--98

EN

The iterative inversion of neural networks has been used in solving problems of adaptive control due to its good performance of information processing. In this paper an iterative inversion neural network with L₂ penalty term has been presented trained by using the classical gradient descent method. We mainly focus on the theoretical analysis of this proposed algorithm such as monotonicity of error function, boundedness of input sequences and weak (strong) convergence behavior. For bounded property of inputs, we rigorously proved that the feasible solutions of input are restricted in a measurable field. The weak convergence means that the gradient of error function with respect to input tends to zero as the iterations go to infinity while the strong convergence stands for the iterative sequence of input vectors convergence to a fixed optimal point.

2

Convergence Analysis of Multilayer Feedforward Networks Trained with Penalty Terms: A review

Wang J., Yang G., Liu S., Zurada J. M.

Journal of Applied Computer Science Methods

|

2015

|

Vol. 7 No. 2

89--103

EN

Gradient descent method is one of the popular methods to train feedforward neural networks. Batch and incremental modes are the two most common methods to practically implement the gradient-based training for such networks. Furthermore, since generalization is an important property and quality criterion of a trained network, pruning algorithms with the addition of regularization terms have been widely used as an efficient way to achieve good generalization. In this paper, we review the convergence property and other performance aspects of recently researched training approaches based on different penalization terms. In addition, we show the smoothing approximation tricks when the penalty term is non-differentiable at origin.

3

Multiplicative Algorithm for Correntropy-Based Nonnegative Matrix Factorization

Hosseini-Asl E., Zurada J. M.

Journal of Applied Computer Science Methods

|

2013

|

Vol. 5 No. 2

89--104

EN

Nonnegative matrix factorization (NMF) is a popular dimension reduction technique used for clustering by extracting latent features from highdimensional data and is widely used for text mining. Several optimization algorithms have been developed for NMF with different cost functions. In this paper we evaluate the correntropy similarity cost function. Correntropy is a nonlinear localized similarity measure which measures the similarity between two random variables using entropy-based criterion, and is especially robust to outliers. Some algorithms based on gradient descent have been used for correntropy cost function, but their convergence is highly dependent on proper initialization and step size and other parameter selection. The proposed general multiplicative factorization algorithm uses the gradient descent algorithm with adaptive step size to maximize the correntropy similarity between the data matrix and its factorization. After devising the algorithm, its performance has been evaluated for document clustering. Results were compared with constrained gradient descent method using steepest descent and L-BFGS methods. The simulations show that the performance of steepest descent and LBFGS convergence are highly dependent on gradient descent step size which depends on σ parameter of correntropy cost function. However, the multiplicative algorithm is shown to be less sensitive to σ parameterand yields better clustering results than other algorithms. The results demonstrate that clustering performance measured by entropy and purity improve the clustering. The multiplicative correntropy-based algorithm also shows less variation in accuracy of document clusters for variable number of clusters. The convergence of each algorithm is also investigated, and the experiments have shown that the multiplicative algorithm converges faster than L-BFGS and steepest descent method.

4

Robust Identification Approach to Individualized Anemia Model

Akabua E., Inanc T., Gaweda A., Brier M. E., Kim S., Zurada J. M.

Journal of Applied Computer Science Methods

|

2011

|

Vol. 3 No. 2

65--75

EN

Over ninety percent of End Stage Renal Disease (ESRD) patients suffer from anemia due to insufficient endogenous production of human erythropoietin. Until the advent of Recombinant Human Erythropoietin (r-HuEPO) over 30 years ago, patients with ESRD were treated mainly with multiple blood transfusions. The high cost of r-HuEPO in addition to the narrow margin between an effective do-sage and toxicity in drug administration calls for optimal dosage strategy capable of minimizing cost and toxicity while at the same time achieving the desired do-sage outcome. It is well known from control theory that a controller can be de-signed for any plant provided there is readily available a valid model for such a plant. We present Robust Identification procedure, a dimensionality reduction technique capable of capturing the inherent dynamics of anemia patients; conse-quently producing individualized model suitable for robust control synthesis and any other controller design methodologies.

5

Searching and Reasoning with Distributed Resources in Computational Intelligence and Machine

Wojtusiak J, Chorowski J, Pietrzykowski J., Zurada J. M.

Journal of Applied Computer Science Methods

|

2009

|

Vol. 1 No. 2

5--19

EN

Virtual organizations (VO) are geographically distributed groups of people sharing common goals and willingness to collaborate. One of the important roles of virtual organizations is to facilitate sharing resources related to the area of collaboration. This paper presents an approach to handling resources in Computational Intelligence and Machine Learning distributed over a large number of sites. Resources are first discovered in the internet, evaluated, and then shared among users. Reasoning and adaptation methods can then be applied to best fit resources into users’ needs without a long search process.

6

International Journal of Applied Computer Science Methods: Authors’ Instructions for the Preparation of the Submission Format of the Final Version of the Manuscript

Zurada J. M., Cader A., Przybyszewski K.

Journal of Applied Computer Science Methods

|

2009

|

Vol. 1 No. 1

93--99

EN

The abstract should summarize the content of the paper and should contain at least 50 and at most 120 words. It should be set in 9-point font size and should be inset 1.0 cm from the right and left margins. Following the abstract, at least three and up to six keywords should be included. The keywords must characterize the subject of the paper and assists in indexing.

7

Collaborative Web-based Workspaces for an Academic Department

Zurada J. M., Habas P A, Muezzinoglu M K

Journal of Applied Computer Science Methods

|

2009

|

Vol. 1 No. 1

25--36

EN

Continuous program improvement and expectations of productivity of academic departments necessitate the use of web-based information sharing. This paper outlines how simple-to-use web-based environments such as documents repositories and collaborative ABET workspaces can improve the workflow of an academic department. They can boost the departmental productivity and faculty involvement in many aspects of teaching and accreditation efforts, and assist with faculty, staff and students interaction. Such environments reduce email communications and can vary by their degree of interactivity. Document repositories can be read-only, while interactive workspaces will typically extend read-write privileges to all participants. Furthermore, collaborative workspaces offer a democratic and open environment that encourages faculty and staff interaction. Finally, such workspaces remain easily scalable for projects of any size, and offer simplicity of navigation and easy-to-organize table of contents. The discussion of web-based tools has been illustrated with several examples of departmental repositories and interactive workspaces. Implementation guidelines are also offered for those interested in custom design of such tools.