Nonnegative matrix factorization (NMF) is primarily a linear dimensionality reduction technique that factorizes a nonnegative data matrix into two smaller nonnegative matrices: one that represents the basis of the new subspace and the second that holds the coefficients of all the data points in that new space. In principle, the nonnegativity constraint forces the representation to be sparse and parts based. Instead of extracting holistic features from the data, real parts are extracted that should be significantly easier to interpret and analyze. The size of the new subspace selects how many features will be extracted from the data. An effective choice should minimize the noise while extracting the key features. We propose a mechanism for selecting the subspace size by using a minimum description length technique. We demonstrate that our technique provides plausible estimates for real data as well as accurately predicting the known size of synthetic data. We provide an implementation of our code in a Matlab format.

This letter considers how a number of modern Markov chain Monte Carlo (MCMC) methods can be applied for parameter estimation and inference in state-space models with point process observations. We quantified the efficiencies of these MCMC methods on synthetic data, and our results suggest that the Reimannian manifold Hamiltonian Monte Carlo method offers the best performance. We further compared such a method with a previously tested variational Bayes method on two experimental data sets. Results indicate similar performance on the large data sets and superior performance on small ones. The work offers an extensive suite of MCMC algorithms evaluated on an important class of models for physiological signal analysis.

We present a variational Bayesian (VB) approach for the state and parameter inference of a state-space model with point-process observations, a physiologically plausible model for signal processing of spike data. We also give the derivation of a variational smoother, as well as an efficient online filtering algorithm, which can also be used to track changes in physiological parameters. The methods are assessed on simulated data, and results are compared to expectation-maximization, as well as Monte Carlo estimation techniques, in order to evaluate the accuracy of the proposed approach. The VB filter is further assessed on a data set of taste-response neural cells, showing that the proposed approach can effectively capture dynamical changes in neural responses in real time.

Physiological signals such as neural spikes and heartbeats are discrete events in time, driven by continuous underlying systems. A recently introduced data-driven model to analyze such a system is a state-space model with point process observations, parameters of which and the underlying state sequence are simultaneously identified in a maximum likelihood setting using the expectation-maximization (EM) algorithm. In this note, we observe some simple convergence properties of such a setting, previously un-noticed. Simulations show that the likelihood is unimodal in the unknown parameters, and hence the EM iterations are always able to find the globally optimal solution.

Population coding is a simplified model of distributed information processing in the brain. This study investigates the performance and implementation of a sequential Bayesian decoding (SBD) paradigm in the framework of population coding. In the first step of decoding, when no prior knowledge is available, maximum likelihood inference is used; the result forms the prior knowledge of stimulus for the second step of decoding. Estimates are propagated sequentially to apply maximum a posteriori (MAP) decoding in which prior knowledge for any step is taken from estimates from the previous step. Not only do we analyze the performance of SBD, obtaining the optimal form of prior knowledge that achieves the best estimation result, but we also investigate its possible biological realization, in the sense that all operations are performed by the dynamics of a recurrent network. In order to achieve MAP, a crucial point is to identify a mechanism that propagates prior knowledge. We find that this could be achieved by short-term adaptation of network weights according to the Hebbian learning rule. Simulation results on both constant and time-varying stimulus support the analysis.

The application of statistical physics to the study of the learning curves of feedforward connectionist networks has to date been concerned mostly with perceptron-like networks. Recent work has extended the theory to networks such as committee machines and parity machines, and an important direction for current and future research is the extension of this body of theory to further connectionist networks. In this article, we use this formalism to investigate the learning curves of gaussian radial basis function networks (RBFNs) having fixed basis functions. (These networks have also been called generalized linear regression models.) We address the problem of learning linear and nonlinear, realizable and unrealizable, target rules from noise-free training examples using a stochastic training algorithm. Expressions for the generalization error, defined as the expected error for a network with a given set of parameters, are derived for general gaussian RBFNs, for which all parameters, including centers and spread parameters, are adaptable. Specializing to the case of RBFNs with fixed basis functions (basis functions having parameters chosen without reference to the training examples), we then study the learning curves for these networks in the limit of high temperature.

The principle of F-projection, in sequential function estimation, provides a theoretical foundation for a class of gaussian radial basis function networks known as the resource allocating networks (RAN). The ad hoc rules for adaptively changing the size of RAN architectures can be justified from a geometric growth criterion defined in the function space. In this paper, we show that the same arguments can be used to arrive at a pruning with replacement rule for RAN architectures with a limited number of units. We illustrate the algorithm on the laser time series prediction problem of the Santa Fe competition and show that results similar to those of the winners of the competition can be obtained with pruning and replacement.

This article addresses the question of whether some recent Vapnik-Chervonenkis (VC) dimension-based bounds on sample complexity can be regarded as a practical design tool. Specifically, we are interested in bounds on the sample complexity for the problem of training a pattern classifier such that we can expect it to perform valid generalization. Early results using the VC dimension, while being extremely powerful, suffered from the fact that their sample complexity predictions were rather impractical. More recent results have begun to improve the situation by attempting to take specific account of the precise algorithm used to train the classifier. We perform a series of experiments based on a task involving the classification of sets of vowel formant frequencies. The results of these experiments indicate that the more recent theories provide sample complexity predictions that are significantly more applicable in practice than those provided by earlier theories; however, we also find that the recent theories still have significant shortcomings.

In this paper, we investigate the problem of optimal sequential learning, viewed as a problem of estimating an underlying function sequentially rather than estimating a set of parameters of the neural network. First, we arrive at a suboptimal solution to the sequential estimate that can be mapped by a growing gaussian radial basis function (GaRBF) network. This network adds hidden units for each observation. The function space approach in which the estimates are represented as vectors in a function space is used in developing a growth criterion to limit its growth. A simplification of the criterion leads to two joint criteria on the distance of the present pattern and the existing unit centers in the input space and on the approximation error of the network for the given observation to be satisfied together. This network is similar to the resource allocating network (RAN) (Platt 1991a) and hence RAN can be interpreted from a function space approach to sequential learning. Second, we present an enhancement to the RAN. The RAN either allocates a new unit based on the novelty of an observation or adapts the network parameters by the LMS algorithm. The function space interpretation of the RAN lends itself to an enhancement of the RAN in which the extended Kalman filter (EKF) algorithm is used in place of the LMS algorithm. The performance of the RAN and the enhanced network are compared in the experimental tasks of function approximation and time-series prediction demonstrating the superior performance of the enhanced network with fewer number of hidden units. The approach adopted here has led us toward the minimal network required for a sequential learning problem.