Joaquin Sitte

The parameterless self-organizing map algorithm

The parameterless self-organizing map (PLSOM) is a new neural network algorithm based on the self-organizing map (SOM). It eliminates the need for a learning rate and annealing schemes for learning rate and neighborhood size. We discuss the relative performance of the PLSOM and the SOM and demonstrate some tasks in which the SOM fails but the PLSOM performs satisfactory. Finally we discuss some example applications of the PLSOM and present a proof of ordering under certain limited conditions.

Adaptive nearest neighbor pattern classification

A variant of nearest-neighbor (NN) pattern classification and supervised learning by learning vector quantization (LVQ) is described. The decision surface mapping method (DSM) is a fast supervised learning algorithm and is a member of the LVQ family of algorithms. A relatively small number of prototypes are selected from a training set of correctly classified samples. The training set is then used to adapt these prototypes to map the decision surface separating the classes. This algorithm is compared with NN pattern classification, learning vector quantization, and a two-layer perceptron trained by error backpropagation. When the class boundaries are sharply defined (i.e., no classification error in the training set), the DSM algorithm outperforms these methods with respect to error rates, learning rates, and the number of prototypes required to describe class boundaries.

A constructive method for multivariate function approximation by multilayer perceptrons

Mathematical theorems establish the existence of feedforward multilayered neural networks, based on neurons with sigmoidal transfer functions, that approximate arbitrarily well any continuous multivariate function. However, these theorems do not provide any hint on how to find the network parameters in practice. It is shown how to construct a perceptron with two hidden layers for multivariate function approximation. Such a network can perform function approximation in the same manner as networks based on Gaussian potential functions, by linear combination of local functions.

A cartpole experiment benchmark for trainable controllers

The inverted pendulum problem, i.e., the cartpole, which is often used for demonstrating the success of neural network learning methods, is addressed. It is shown that a random search in weight space can quickly uncover coefficients (weights) for controllers that work over a wide range of initial conditions. This result indicates that success in finding a satisfactory neural controller is not sufficient proof for the effectiveness of unsupervised training methods. By analyzing the dynamics of the linear controller, the cartpole problem is reformulated to make it a more stringent test for neural training methods. A review of the literature on unsupervised training methods for cartpole controllers shows that the published results are difficult to compare and that for most of the methods there is not clear evidence of better performance than the random search method.<<ETX>>

Neural Networks Approach to the Random Walk Dilemma of Financial Time Series

Predictions of financial time series often show a characteristic one step shift relative to the original data as in a random walk. This has been the cause for opposing views whether such time series do contain information that can be extracted for predictions, or are simply random walks. In this case study, we show that NNs that are capable of extracting weak low frequency periodic signals buried in a strong high frequency signal, consistently predict the next value in the series to be the current value, as in a random walk, when used for one-step-ahead predictions of the detrended S&P 500 time series. In particular for the Time Delay Feed Forward Networks and Elman Networks of various configurations, our study supports the view of the detrended S&P 500 being a random walk series. This is consistent with the long standing hypothesis that some financial time series are random walk series.

Analysis of the predictive ability of time delay neural networks applied to the S&P 500 time series

Reported work on financial time series prediction using neural networks often shows a characteristic one step shift relative to the original data. This seems to imply a failure of the neural network (NN), because a shift corresponds to a random walk prediction. Our systematic analysis of different time delay neural networks predictors applied to the detrended S&P 500 time series, indicates that this prediction behavior is not a limitation of the network, but may be a characteristic of the time series. This suggests that there are no short-term correlations in this stockmarket time series, which is consistent with conventional statistical analysis.

Local cluster neural net: Architecture, training and applications

Abstract This paper describes the structure, training and computational abilities of the local cluster (LC) artificial neural net architecture. LC nets are a special class of multilayer perceptrons that use sigmoid functions to generate localised functions. LC nets train as fast as radial basis functions nets and are more general. They are well suited for both, multi-dimensional function approximation and discrete classification. The LC net is the result of our search for a widely applicable neural net architecture suitable for low-cost hardware realisation. The LC net seem particularly well suited for analog VLSI realisation of small-size, low-power, fully parallel neural net chip for real time control applications.

Sound source localisation through active audition

This paper presents a novel method for enabling a robot to determine the direction to a sound source through interacting with its environment. The method uses a new neural network, the parameter-less self-organizing map algorithm, and reinforcement learning to achieve rapid and accurate response.

Progress in Robotics

This volume is a selection of papers of six international conferences that are held under the umbrella of the 12th FIRA RoboWorld congress, in Incheon, Korea, August 16-18, 2009. From the 115 contributed papers 44 papers are included in the volume, which is organized into 6 sections: humanoid robotics, human robot interaction, education and entertainment, cooperative robotics, robotic system design, and learning, optimization, communication. The volume is intended to provide readers with the recent technical progresses in robotics, human robot interactions, cooperative robotics and the related fields.

Demand-Compliant Design

In this paper, we describe, in detail, a design method that assures that the designed product satisfies a set of prescribed demands while, at the same time, providing a concise representation of the design that facilitates communication in multidisciplinary design teams. This Demand Compliant Design (DeCoDe) method was in itself designed to comply with a set of demands. The demands on the method were determined by an analysis of some of the most widely used design methods and from the needs arising in the practice of design for quality. We show several modes of use of the DeCoDe method and illustrate with examples.