It is well-known that the so-called non-uniqueness problem appears in adaptive filters for stereo acoustic echo cancelers and one of the effective solutions is the input-sliding method that utilizes a time-varying preprocessor. In this paper, the convergence speed of the input-sliding method is examined. For adaptive algorithms based on orthogonal projection, such as the normalized LMS algorithm, the convergence speed is maximized when the learning coefficient is unity since the orthogonal projection is completed then. However, if the input signal vectors do not span the whole space but exist in a subspace which varies in time, then the learning coefficient which achieves the fastest convergence is found to be not unity but between one and two. The geometrical consideration of the tap-weight vector explains the phenomena and the theoretical results agree well with those of computer simulations.
In this paper, a method of finding a local minimum is examined for LQ optimal control problem of constant output feedback. A necessary condition for optimality is given as matrix algebraic equations for a modified LQ performance index, and a method of tracing a solution path from a state feedback gain to an output feedback gain is proposed. In order to trace the solution curve that may have branch, a new homotopy method based on series expansion of approximate algebra is applied. Initial gains are obtained from not only a positive definite solution but also the other solutions of Riccati equation, then it is examined numerically whether more than one local minimum can be obtained by tracing the solutions starting from the initial gains. A numerical example which has two local minima shows that one local minimum is connected to the optimal state feedback gain, but the other is not connected to any gains given by the solutions of Riccati equation.
Reactive scheduling approach is considered effective in manufacturing systems, which undergoes various types of delays due to disruption such as machine breakdowns, urgent jobs and so forth. This paper deal with a problem in what situation we should conduct schedule revision under reactive scheduling to obtain an efficient revised schedule. We propose a new model for reactive scheduling based on a control limit policy idea with critical number of delayed tasks which can be a measure to determine the suitable timing of schedule revision. Some computational experiments demonstrate the effectiveness of the proposed model by applying it to job shop scheduling problems with due date.
Web search engines are useful tools which can meet the various information needs of users. However they often return hit-lists which contain many unnecessary pages. This paper proposes a method which automatically removes those unnecessary pages by learning filters through relevance feedback. Filters consist of several rules, each of which describes conditions for discriminating relevant pages using useful cooccurences or proximities among words, and parts in a page where those conditions are applicable. Thus they enable to classify a variety of pages precisely. Moreover, users are free to generate and apply filters at anytime. Through experiments we demonstarate that our filters increase the number of relevant pages we can get in a retrieval, compared to representatives of web search engine and relevance feedback method.
Along the development of computer networks, the necessity of authenticating a person by biometrics is increasing. In this paper, we propose a method for matching of two online signatures. For online signature verification problems, it is necessary to assume that only one or very small number of real signatures can be used because of the property of the problem. We apply DP (Dynamic Programming) matching, where one signature is to be scaled and shifted by time-variant model. We propose a method to apply DP matching and the parameter estimation scheme interchangeably. Experiments using real data validate this method.
The continuation method is one approach to solve the system of nonlinear equations. For variational inequality problems, interior point like continuation methods based on the KKT conditions for VIP have been proposed. Kanzow and Jiang considered one parameter continuation method for strongly monotone variational inequality problems and showed that a continuation path exists under the linear independence constraint qualification condition. In this paper, we show that, the unique continuation path approaching to a solution exists for strongly monotone problems under the Slater's constraint qualification instead of the linear independence constraint qualification. Moreover, this theorem is shown to give another proof of the existence theorem for a continuation path of Chen and Harker.
Non-linear extensions of Principal Component Analysis (PCA) have been developed for detecting the lower-dimensional representations of real world data sets. Fuzzy c-Varieties (FCV) is the linear fuzzy clustering algorithm that can be regarded as a Local PCA technique. However least squares techniques often fail to account for “outliers”. This paper proposes a technique for making the FCV algorithm robust to intra-sample outliers. The objective function based on the lower rank approximation of the data matrix is minimized by a robust M-estimation algorithm that is similar to FCM-type iterative procedures. The new method is also useful for estimating missing values and a numerical experiment of Collaborative Filtering reveals an improvement in recommendation performance.