Abstract: The conjugate gradient optimization algorithm
usually used for nonlinear least squares is presented and is
combined with the modified back propagation algorithm yielding
a new fast training multilayer perceptron (MLP) algorithm
(CGFR/AG). The approaches presented in the paper consist of
three steps: (1) Modification on standard back propagation
algorithm by introducing gain variation term of the activation
function, (2) Calculating the gradient descent on error with
respect to the weights and gains values and (3) the determination
of the new search direction by exploiting the information
calculated by gradient descent in step (2) as well as the previous
search direction. The proposed method improved the training
efficiency of back propagation algorithm by adaptively modifying
the initial search direction. Performance of the proposed method
is demonstrated by comparing to the conjugate gradient algorithm
from neural network toolbox for the chosen benchmark. The
results show that the number of iterations required by the
proposed method to converge is less than 20% of what is required
by the standard conjugate gradient and neural network toolbox
algorithm.
Abstract: This paper introduces new algorithms (Fuzzy relative
of the CLARANS algorithm FCLARANS and Fuzzy c Medoids
based on randomized search FCMRANS) for fuzzy clustering of
relational data. Unlike existing fuzzy c-medoids algorithm (FCMdd)
in which the within cluster dissimilarity of each cluster is minimized
in each iteration by recomputing new medoids given current
memberships, FCLARANS minimizes the same objective function
minimized by FCMdd by changing current medoids in such away
that that the sum of the within cluster dissimilarities is minimized.
Computing new medoids may be effected by noise because outliers
may join the computation of medoids while the choice of medoids in
FCLARANS is dictated by the location of a predominant fraction of
points inside a cluster and, therefore, it is less sensitive to the
presence of outliers. In FCMRANS the step of computing new
medoids in FCMdd is modified to be based on randomized search.
Furthermore, a new initialization procedure is developed that add
randomness to the initialization procedure used with FCMdd. Both
FCLARANS and FCMRANS are compared with the robust and
linearized version of fuzzy c-medoids (RFCMdd). Experimental
results with different samples of the Reuter-21578, Newsgroups
(20NG) and generated datasets with noise show that FCLARANS is
more robust than both RFCMdd and FCMRANS. Finally, both
FCMRANS and FCLARANS are more efficient and their outputs
are almost the same as that of RFCMdd in terms of classification
rate.
Abstract: Installation of power compensation equipment in
some cases places additional buses into the system. Therefore, a total
number of power flow equations and voltage unknowns increase due
to additional locations of installed devices. In this circumstance, power flow calculation is more complicated. It may result in a
computational convergence problem. This paper presents a power flow calculation by using Newton-Raphson iterative method together
with the proposed load transfer technique. This concept is to eliminate additional buses by transferring installed loads at the new buses to existing two adjacent buses. Thus, the total number of power
flow equations is not changed. The overall computational speed is
expectedly shorter than that of solving the problem without applying the load transfer technique. A 15-bus test system is employed for test
to evaluate the effectiveness of the proposed load transfer technique. As a result, the total number of iteration required and execution time
is significantly reduced.
Abstract: A novel typical day prediction model have been built and validated by the measured data of a grid-connected solar photovoltaic (PV) system in Macau. Unlike conventional statistical method used by previous study on PV systems which get results by averaging nearby continuous points, the present typical day statistical method obtain the value at every minute in a typical day by averaging discontinuous points at the same minute in different days. This typical day statistical method based on discontinuous point averaging makes it possible for us to obtain the Gaussian shape dynamical distributions for solar irradiance and output power in a yearly or monthly typical day. Based on the yearly typical day statistical analysis results, the maximum possible accumulated output energy in a year with on site climate conditions and the corresponding optimal PV system running time are obtained. Periodic Gaussian shape prediction models for solar irradiance, output energy and system energy efficiency have been built and their coefficients have been determined based on the yearly, maximum and minimum monthly typical day Gaussian distribution parameters, which are obtained from iterations for minimum Root Mean Squared Deviation (RMSD). With the present model, the dynamical effects due to time difference in a day are kept and the day to day uncertainty due to weather changing are smoothed but still included. The periodic Gaussian shape correlations for solar irradiance, output power and system energy efficiency have been compared favorably with data of the PV system in Macau and proved to be an improvement than previous models.
Abstract: In this paper, we propose a single sample path based
algorithm with state aggregation to optimize the average rewards of
singularly perturbed Markov reward processes (SPMRPs) with a
large scale state spaces. It is assumed that such a reward process
depend on a set of parameters. Differing from the other kinds of
Markov chain, SPMRPs have their own hierarchical structure. Based
on this special structure, our algorithm can alleviate the load in the
optimization for performance. Moreover, our method can be applied
on line because of its evolution with the sample path simulated.
Compared with the original algorithm applied on these problems of
general MRPs, a new gradient formula for average reward
performance metric in SPMRPs is brought in, which will be proved
in Appendix, and then based on these gradients, the schedule of the
iteration algorithm is presented, which is based on a single sample
path, and eventually a special case in which parameters only
dominate the disturbance matrices will be analyzed, and a precise
comparison with be displayed between our algorithm with the old
ones which is aim to solve these problems in general Markov reward
processes. When applied in SPMRPs, our method will approach a fast
pace in these cases. Furthermore, to illustrate the practical value of
SPMRPs, a simple example in multiple programming in computer
systems will be listed and simulated. Corresponding to some practical
model, physical meanings of SPMRPs in networks of queues will be
clarified.