Targeting c-Myc-activated genes with a correlation method: detection of global changes in large gene expression network dynamics.

This work studies the dynamics of a gene expression time series network. The network, which is obtained from the correlation of gene expressions, exhibits global dynamic properties that emerge after a cell state perturbation. The main features of this network appear to be more robust when compared with those obtained with a network obtained from a linear Markov model. In particular, the network properties strongly depend on the exact time sequence relationships between genes and are destroyed by random temporal data shuffling. We discuss in detail the problem of finding targets of the c-myc protooncogene, which encodes a transcriptional regulator whose inappropriate expression has been correlated with a wide array of malignancies. The data used for network construction are a time series of gene expression, collected by microarray analysis of a rat fibroblast cell line expressing a conditional Myc-estrogen receptor oncoprotein. We show that the correlation-based model can establish a clear relationship between network structure and the cascade of c-myc-activated genes.

The effect of noise on a class of energy-based learning rules.

Westudy the selectivity properties of neurons based on BCM and kurtosis energy functions in a general case of noisy high-dimensional input space. The proposed approach, which is used for characterization of the stable states, can be generalized to a whole class of energy functions. We characterize the critical noise levels beyond which the selectivity is destroyed. We also perform a quantitative analysis of such transitions, which shows interesting dependency on data set size. We observe that the robustness to noise of the BCM neuron (Bienenstock, Cooper, & Munro, 1982; Intrator & Cooper, 1992) increases as a function of dimensionality. We explicitly compute the separability limit of BCM and kurtosis learning rules in the case of a bimodal input distribution. Numerical simulations show a stronger robustness of the BCM rule for practical data set size when compared with kurtosis.

(Coarse coding of shape fragments) + (retinotopy) approximately = representation of structure.

The ability to deal with object structure--to determine what is where in a given object, rather than merely to categorize or identify it--has been hitherto considered the prerogative of 'structural description' approaches, which represent shapes as categorical compositions of generic parts taken from a small alphabet. In this note, we propose a simple extension to a theoretically motivated and extensively tested appearance-based model of recognition and categorization, which should make it capable of representing object structure. We describe a pilot implementation of the extended model, survey independent evidence supporting its modus operandi, and outline a research program focused on achieving a range of object processing capabilities, including reasoning about structure, within a unified appearance-based framework.

Solutions of the BCM learning rule in a network of lateral interacting nonlinear neurons.

We introduce a new method for obtaining the fixed points for neurons that follow the BCM learning rule. The new formalism, which is based on the objective function formulation, permits analysis of a laterally connected network of nonlinear neurons and allows explicit calculation of the fixed points under various network conditions. We show that the stable fixed points, in terms of the postsynaptic activity, are not altered by the lateral connectivity or nonlinearity. We show that the lateral connectivity alters the probability of attaining different states in a network of interacting neurons. We further show the exact alteration in presynaptic weights as a result of the neuronal nonlinearity.

Boosted mixture of experts: an ensemble learning scheme.

We present a new supervised learning procedure for ensemble machines, in which outputs of predictors, trained on different distributions, are combined by a dynamic classifier combination model. This procedure may be viewed as either a version of mixture of experts (Jacobs, Jordan, Nowlan, & Hintnon, 1991), applied to classification, or a variant of the boosting algorithm (Schapire, 1990). As a variant of the mixture of experts, it can be made appropriate for general classification and regression problems by initializing the partition of the data set to different experts in a boostlike manner. If viewed as a variant of the boosting algorithm, its main gain is the use of a dynamic combination model for the outputs of the networks. Results are demonstrated on a synthetic example and a digit recognition task from the NIST database and compared with classifical ensemble approaches.

Boosting regression estimators.

There is interest in extending the boosting algorithm (Schapire, 1990) to fit a wide range of regression problems. The threshold-based boosting algorithm for regression used an analogy between classification errors and big errors in regression. We focus on the practical aspects of this algorithm and compare it to other attempts to extend boosting to regression. The practical capabilities of this model are demonstrated on the laser data from the Santa Fe times-series competition and the Mackey-Glass time series, where the results surpass those of standard ensemble average.

Receptive Field Formation in Natural Scene Environments. Comparison of Single-Cell Learning Rules.

We study several statistically and biologically motivated learning rules using the same visual environment: one made up of natural scenes and the same single-cell neuronal architecture. This allows us to concentrate on the feature extraction and neuronal coding properties of these rules. Included in these rules are kurtosis and skewness maximization, the quadratic form of the Bienenstock-Cooper-Munro (BCM) learning rule, and single-cell independent component analysis. Using a structure removal method, we demonstrate that receptive fields developed using these rules depend on a small portion of the distribution. We find that the quadratic form of the BCM rule behaves in a manner similar to a kurtosis maximization rule when the distribution contains kurtotic directions, although the BCM modification equations are computationally simpler.

Conservation of total T-cell counts during HIV infection: alternative hypotheses and implications.

While CD4+ T-cell counts in the blood of HIV-infected individuals gradually decrease, there is a parallel increase in the number of blood CD8+ T cells such that the total number of T cells remains essentially constant for several years (1). The basis and significance of this phenomenon are not known. Based on a statistical analysis of longitudinal T-cell counts from the Transfusion Safety Study (TSS) database and on theoretical considerations, we evaluate several alternative models, including versions of the "blind homeostasis" (BH) hypothesis (1-3). At issue is the nature of the homeostatic regulation of lymphocytes and its apparent failure in HIV infection. The most plausible explanation for the conservation of total blood T-cell numbers while subset ratios change is that CD4+ and CD8+ T cells compete for a limited access to the blood compartment. Such interaction between the subsets implies, in particular, that changes in the number of CD4+ T cells occurring in other tissues cannot be reliably inferred from those observed in the blood. We reiterate propositions made earlier (4) that much of the apparent "depletion" of CD4+ lymphocytes during the asymptomatic phase of HIV infection may be attributed to redistribution between the tissues and the blood compartment.

Multimodality exploration by an unsupervised projection pursuit neural network.

Graphical inspection of multimodality is demonstrated using unsupervised lateral-inhibition neural networks. Three projection pursuit indexes are compared on low-dimensional simulated and real-world data: principal components, Legendre polynomial, and projection pursuit network.

Competitive learning in biological and artificial neural computation.

In this review we will briefly discuss 'classical' competitive learning and approaches to competitive learning that involve mixtures of experts. We will then focus on competitive learning that is guided by the temporal structure that is present within the stimuli. In this context, we will describe a general principle for resource allocation and memory management, that may account for a range of psychophysical and neurophysiological findings.

BCM network develops orientation selectivity and ocular dominance in natural scene environment.

A two-eye visual environment is used in training a network of BCM neurons. We study the effect of misalignment between the synaptic density functions from the two eyes, on the formation of orientation selectivity and ocular dominance in a lateral inhibition network. The visual environment we use is composed of natural images. We show that for the BCM rule a natural image environment with binocular cortical misalignment is sufficient for producing networks with orientation-selective cells and ocular dominance columns. This work is an extension of our previous single cell misalignment model Shouval et al., 1996.

Effect of binocular cortical misalignment on ocular dominance and orientation selectivity.

We model a two-eye visual environment composed of natural images and study its effect on single cell synaptic modification. In particular, we study the effect of binocular cortical misalignment on receptive field formation after eye opening. We show that binocular misalignment affects principal component analysis (PCA) and Bienenstock, Cooper, and Munro (BCM) learning in different ways. For the BCM learning rule this misalignment is sufficient to produce varying degrees of ocular dominance, whereas for PCA learning binocular neurons emerge in every case.

A model of prenatal acquisition of speech parameters.

An unsupervised neural network model inductively acquires the ability to distinguish categorically the stop consonants of English, in a manner consistent with prenatal and early postnatal auditory experience, and without reference to any specialized knowledge of linguistic structure or the properties of speech. This argues against the common assumption that linguistic knowledge, and speech perception in particular, cannot be learned and must therefore be innately specified.

A novel cardiotoxic polypeptide from the venom of Atractaspis engaddensis (burrowing asp): cardiac effects in mice and isolated rat and human heart preparations.

A new cardiotoxic polypeptide isolated from the venom of the snake Atractaspis engaddensis has an LD50 of 15 micrograms/kg body weight in white mice. Intravenous administration in mice of lethal doses of the toxin causes, within seconds, marked changes in the ECG, consisting primarily of a transient slope elevation of the S-T segment, a temporary diminution of the S-wave and an increase in the amplitudes of the R- and T-waves. Concomitantly, and apparently unrelated to these changes, a severe A-V block develops and leads to complete cardiac arrest within a few min. Studies with rat and human isolated heart preparations showed that the toxin exerts a powerful coronary vasoconstriction (rats), and positive inotropic effects (rats and humans).