Vladimir Ceperic

A Strategy for Short-Term Load Forecasting by Support Vector Regression Machines

This paper presents a generic strategy for short-term load forecasting (STLF) based on the support vector regression machines (SVR). Two important improvements to the SVR based load forecasting method are introduced, i.e., procedure for generation of model inputs and subsequent model input selection using feature selection algorithms. One of the objectives of the proposed strategy is to reduce the operator interaction in the model-building procedure. The proposed use of feature selection algorithms for automatic model input selection and the use of the particle swarm global optimization based technique for the optimization of SVR hyper-parameters reduces the operator interaction. To confirm the effectiveness of the proposed modeling strategy, the model has been trained and tested on two publicly available and well-known load forecasting data sets and compared to the state-of-the-art STLF algorithms yielding improved accuracy.

Design and Optimization of Self-Biased Complementary Folded Cascode

This paper presents design and optimization procedure of a self-biased complementary folded cascade. A self-biased scheme is chosen as a technique that saves power and circuit area, and is less sensitive to process variations. The gain of basic folded cascode is enhanced using a gain boosting approach based on common source self-biased amplifiers. The circuits are optimized using the global optimization approach with the cost function calculated by circuit simulations. The hybrid approach to optimization is used combining the global search strategy using particle swarm optimization (PSO) and direct pattern search (DPS) method used as local search strategy. A complementary folded cascode operational amplifier is designed in the 0.35 mum CMOS technology with the 3.3 V power supply voltage

Recurrent sparse support vector regression machines trained by active learning in the time-domain

A method for the sparse solution of recurrent support vector regression machines is presented. The proposed method achieves a high accuracy versus complexity and allows the user to adjust the complexity of the resulting model. The sparse representation is guaranteed by limiting the number of training data points for the support vector regression method. Each training data point is selected based on the accuracy of the fully recurrent model using the active learning principle applied to the successive time-domain data. The user can adjust the training time by selecting how often the hyper-parameters of the algorithm should be optimised. The advantages of the proposed method are illustrated on several examples, and the experiments clearly show that it is possible to reduce the number of support vectors and to significantly improve the accuracy versus complexity of recurrent support vector regression machines.

Modeling of analog circuits by using support vector regression machines

The support vector regression method is used for modeling of electronic circuits. The method ensures simple, robust and accurate modeling of electronic circuits. It yields very good results for situations not specified in the learning data set, demonstrating very good generalization property of support vector machines. The method is applicable to modeling based on the measurements or device-level circuit simulations. Several GaAs circuits (buffer, resistive mixer, ring oscillator) are modeled using the proposed method.

Modelling of Electromagnetic Immunity of Integrated Circuits by Artificial Neural Networks

A simple and efficient method of modelling electromagnetic immunity (EMI) of integrated circuits (IC) with respect to conducted electromagnetic interference by using artificial neural networks (ANN) is presented. A pulse signal generator is controlled by an ANN to improve stability, robustness and accuracy of the model. A simple and effective way to obtain necessary data for learning the artificial neural network used in EMI modelling is presented. The methodology described in this paper ensures a simple, fast and accurate modelling. As a test case, the EMI model of conducted interference of a simple local interconnect network (LIN) interface circuit is presented.

Reducing Complexity of Echo State Networks with Sparse Linear Regression Algorithms

In this paper the use of sparse linear regression algorithms in echo state networks (ESN) is presented for reducing the number of readouts and improving the robustness and generalization properties of ESNs. Three data sets with overall 80 tests are used to validate the use of sparse linear regression algorithms for echo state networks. It is shown that it is possible to increase accuracy on the test data sets, not used in the ESN training phase, and in the same time reduce the overall number of the required readouts when compared to the standard approach of using ridge linear regression on the echo state network readouts.

Sparse multikernel support vector regression machines trained by active learning

A method for the sparse multikernel support vector regression machines is presented. The proposed method achieves a high accuracy versus complexity ratio and allows the user to adjust the complexity of the resulting models. The sparse representation is guaranteed by limiting the number of training data points for the support vector regression method. Each training data point is selected based on its influence on the accuracy of the model using the active learning principle. A different kernel function is attributed to each training data point, yielding multikernel regressor. The advantages of the proposed method are illustrated on several examples and the experiments show the advantages of the proposed method.

Sparse ε-tube support vector regression by active learning

A method for the sparse solution of

Black-Box Modelling of AC-DC Rectifiers for RFID Applications Using Support Vector Regression Machines

\varepsilon