An usual task in large data set analysis is searching for an appropriate data representation in a space of fewer dimensions. One of the most efficient methods to solve this task is factor analysis. In this paper, we compare seven methods for Boolean factor analysis (BFA) in solving the so-called bars problem (BP), which is a BFA benchmark. The performance of the methods is evaluated by means of information gain. Study of the results obtained in solving BP of different levels of complexity has allowed us to reveal strengths and weaknesses of these methods. It is shown that the Likelihood maximization Attractor Neural Network with Increasing Activity (LANNIA) is the most efficient BFA method in solving BP in many cases. Efficacy of the LANNIA method is also shown, when applied to the real data from the Kyoto Encyclopedia of Genes and Genomes database, which contains full genome sequencing for 1368 organisms, and to text data set R52 (from Reuters 21578) typically used for label categorization.

Studied are differences of two approaches to binary data dimension reduction. The first one is Boolean Matrix Factorization and the second one is Expectation Maximization Boolean Factor Analysis. The two BMF methods are used for comparison. First is M8 method from the BMDP statistical software package. The second is the BMF method, as suggested by Belohlavek and Vychodil (BVA2). These two are compared to Expectation Maximization Boolean Factor Analysis extended with binarization step developed here. Generated (Bars problem) and mushroom dataset are used for(experiments. In particular, under scrutiny was the reconstruction ability of the computed factors and the information gain as the measure of dimension reduction. In addition, presented are some general remarks on all the methods being compared.

Methods for hidden structure of high-dimensional binary data discovery are one of the most important challenges facing machine learning community researchers. There are many approaches in literature that try to solve this hitherto rather ill-defined task. In the present study, we propose a most general generative model of binary data for Boolean factor analysis and introduce new Expectation-Maximization Boolean Factor Analysis algorithm which maximizes likelihood of Boolean Factor Analysis solution. Using the so-called bars problem benchmark, we compare efficiencies of Expectation-Maximization Boolean Factor Analysis algorithm with Dendritic Inhibition neural network. Then we discuss advantages and disadvantages of both approaches as regards results quality and methods efficiency.

This letter presents a novel unsupervised sensory matching learning technique for the development of an internal representation of three-dimensional information. The representation is invariant with respect to the sensory modalities involved. Acquisition of the internal representation is demonstrated with a neural network model of a sensorimotor system of a simple model creature, consisting of a tactile-sensitive body and a multiple-degrees-of-freedom arm with proprioceptive sensitivity. Acquisition of the 3D representation as well as a distributed representation of the body scheme, occurs through sensorimotor interactions (i.e., the sensory-motor experience of the creature). Convergence of the learning is demonstrated through computer simulations for the model creature with a 7-DoF arm and a spherical body covered by 20 tactile fields.

The objective of this paper is to introduce a neural-network-based algorithm for word clustering as an extension of the neural-network-based Boolean factor analysis algorithm (Frolov , 2007). It is shown that this extended algorithm supports even the more complex model of signals that are supposed to be related to textual documents. It is hypothesized that every topic in textual data is characterized by a set of words which coherently appear in documents dedicated to a given topic. The appearance of each word in a document is coded by the activity of a particular neuron. In accordance with the Hebbian learning rule implemented in the network, sets of coherently appearing words (treated as factors) create tightly connected groups of neurons, hence, revealing them as attractors of the network dynamics. The found factors are eliminated from the network memory by the Hebbian unlearning rule facilitating the search of other factors. Topics related to the found sets of words can be identified based on the words' semantics. To make the method complete, a special technique based on a Bayesian procedure has been developed for the following purposes: first, to provide a complete description of factors in terms of component probability, and second, to enhance the accuracy of classification of signals to determine whether it contains the factor. Since it is assumed that every word may possibly contribute to several topics, the proposed method might be related to the method of fuzzy clustering. In this paper, we show that the results of Boolean factor analysis and fuzzy clustering are not contradictory, but complementary. To demonstrate the capabilities of this attempt, the method is applied to two types of textual data on neural networks in two different languages. The obtained topics and corresponding words are at a good level of agreement despite the fact that identical topics in Russian and English conferences contain different sets of keywords.

In this paper, we compare three methods for grouping of binary variables: neural network Boolean factor analysis, hierarchical clustering, and a linear factor analysis on the mushroom dataset. In contrast to the latter two traditional methods, the advantage of neural network Boolean factor analysis is its ability to reveal overlapping classes in the dataset. It is shown that the mushroom dataset provides a good demonstration of this advantage because it contains both disjunctive and overlapping classes.

A common problem encountered in disciplines such as statistics, data analysis, signal processing, textual data representation, and neural network research, is finding a suitable representation of the data in the lower dimension space. One of the principles used for this reason is a factor analysis. In this paper, we show that Hebbian learning and a Hopfield-like neural network could be used for a natural procedure for Boolean factor analysis. To ensure efficient Boolean factor analysis, we propose our original modification not only of Hopfield network architecture but also its dynamics as well. In this paper, we describe neural network implementation of the Boolean factor analysis method. We show the advantages of our Hopfield-like network modification step by step on artificially generated data. At the end, we show the efficiency of the method on artificial data containing a known list of factors. Our approach has the advantage of being able to analyze very large data sets while preserving the nature of the data.

The sparse encoded Hopfield like neural network is modified to provide the Boolean factor analysis. New, more efficient method of sequential factor extraction, based on the characteristics behavior of the Lyapunov function is introduced. Efficiency of this attempt is shown not only on simulated data but on real data from Russian parliament but as well.

New neural network approach to the control of arbitrary nonlinear plant is presented. The approach is based on the model of cerebro-cerebellar interaction during the learning of new visuo-motor transformation. The model in turn is based on four main prerequisites: the Marr-Albus-Ito theory of cerebellar learning, the equilibrium point theory of motor control, the columnar organization of the cerebral cortex and the differential neurocontroller theory. The convergence of the learning procedure is proved for the linear plant. Its convergence for nonlinear plants are confirmed by computer simulation of arm reaching movements in the horizontal plane.

Convergence time in Hopfield attractor neural network with parallel dynamics is investigated by computer simulation of networks of extremely large size (up to the number of neurons N=10/sup 5/). A special algorithm is used to avoid storage in the computer memory of both connection matrix and the set of stored prototypes. Thus, the size of the simulated network is restricted only by the processing time. It is shown that asymptotically for N/spl rarr//spl infin/ the number of time steps S which are required to reach the attractor in the vicinity of the recalled prototype is proportional to N/sup /spl gamma// where power index /spl gamma//spl Lt/1.

An intrinsic factor analysis (factorization) framework for information redundancy elimination by means of Hebbian learning in sparsely encoded Hopfield-like neural networks is presented. Computer simulations revealed that the information redundancy, which can be eliminated by factorization, is sparseness dependent. Due to strong similarity of Hopfield-like neural networks to that of CA3 field of the hippocampus and following Marr's ideas we propose a physiological mechanism for redundancy elimination (fractorization) in CA3 and further replay to neocortex in the form of "classificatory units".