Unsupervised learning and temporal context to recall complex robot trajectories

An unsupervised neural network is proposed to learn and recall complex robot trajectories. Two cases are considered: (i) A single trajectory in which a particular arm configuration (state) may occur more than once, and (ii) trajectories sharing states with each other. Ambiguities occur in both cases during recall of such trajectories. The proposed model consists of two groups of synaptic weights trained by competitive and Hebbian learning laws. They are responsible for encoding spatial and temporal features of the input sequences, respectively. Three mechanisms allow the network to deal with repeated or shared states: local and global context units, neurons disabled from learning, and redundancy. The network reproduces the current and the next state of the learned sequences and is able to resolve ambiguities. The model was simulated over various sets of robot trajectories in order to evaluate learning and recall, trajectory sampling effects and robustness.     

Temporal Sequence Compression by an Integrate-and-Fire Model of Hippocampal Area CA3

Cells in the rat hippocampus fire as a function of the animal's location in space. Thus, a rat moving through the world produces a statistically reproducible sequence of place cell firings. With this perspective, spatial navigation can be viewed as a sequence learning problem for the hippocampus. That is, learning entails associating the relationships among a sequence of places that are represented by a sequence of place cell firing. Recent experiments by McNaughton and colleagues suggest the hippocampus can recall a sequence of place cell firings at a faster rate than it was experienced. This speedup, which occurs during slow-wave sleep, is called temporal compression. Here, we show that a simplified model of hippocampal area CA3, based on integrate-and-fire cells and unsupervised Hebbian learning, reproduces this temporal compression. The amount of compression is proportional to the activity level during recall and to the relative timespan of associativity during learning. Compression seems to arise from an alteration of network dynamics between learning and recall. During learning, the dynamics are paced by external input and slowed by a low overall level of activity. During recall, however, external input is absent, and the dynamics are controlled by intrinsic network properties. Raising the activity level by lowering inhibition increases the rate at which the network can transition between previously learned states and thereby produces temporal compression. The tendency for speeding up future activations, however, is limited by the temporal range of associations that were present during learning.     

Modeling and production of robot trajectories using the Temporal Parametrized Self Organizing Maps

In this paper we proposed an unsupervised neural architecture, called Temporal Parametrized Self Organizing Map (TEPSOM), capable of learning and reproducing complex robot trajectories and interpolating new states between the learned ones. The TEPSOM combines the Self-Organizing NARX (SONARX) network, responsible for coding the temporal associations of the robotic trajectory, with the Parametrized Self-Organizing (PSOM) network, responsible for an efficient interpolation mechanism acting on the SONARX neurons. The TEPSOM network is used to model the inverse kinematics of the PUMA 560 robot during the execution of trajectories with repeated states. Simulation results show that the TEPSOM is more accurate than the SONARX in the reproduction of the learned trajectories.     

Dynamic network rewiring determines temporal regulatory functions in Drosophila melanogaster development processes

The identification of network motifs has been widely considered as a significant step towards uncovering the design principles of biomolecular regulatory networks. To date, time‐invariant networks have been considered. However, such approaches cannot be used to reveal time‐specific biological traits due to the dynamic nature of biological systems, and hence may not be applicable to development, where temporal regulation of gene expression is an indispensable characteristic. We propose a concept of a “temporal sequence of network motifs”, a sequence of network motifs in active sub‐networks constructed over time, and investigate significant network motifs in the active temporal sub‐networks of Drosophila melanogaster . Based on this concept, we find a temporal sequence of network motifs which changes according to developmental stages and thereby cannot be identified from the whole static network. Moreover, we show that the temporal sequence of network motifs corresponding to each developmental stage can be used to describe pivotal developmental events.     

Dynamic graph convolution neural network based on spatial-temporal correlation for air quality prediction

Air pollution is a serious threat to both the ecological environment and the physical health of individuals. Therefore, accurate air quality prediction is urgent and necessary for pollution mitigation and residents’ travel. However, few existing models are established based on the dynamic spatiotemporal correlation of air pollutants to predict air quality. In this paper, a novel deep learning model combining the dynamic graph convolutional network and the multi-channel temporal convolutional network (DGC-MTCN) is proposed for air quality prediction. To efficiently represent the time-varying spatial dependencies, a new spatiotemporal dynamic correlation calculation method based on gray relation analysis is proposed to construct dynamic adjacency matrices. Then, the spatiotemporal features are sufficiently extracted by the graph convolutional network and the multi-channel temporal convolutional network. Two real-world air quality datasets collected from Beijing and Fushun are applied to verify the performance of our proposed model. The experimental results show that compared with other baselines, the DGC-MTCN model has excellent prediction accuracy. Especially for the prediction of multi-step and different stations, our model performs better temporal stability and generalization ability.     

Uncovering the synchronization dynamics from correlated neuronal activity quantifies assembly formation

 Synchronous network excitation is believed to play an outstanding role in neuronal information processing. Due to the stochastic nature of the contributing neurons, however, those synchronized states are difficult to detect in electrode recordings. We present a framework and a model for the identification of such network states and of their dynamics in a specific experimental situation. Our approach operationalizes the notion of neuronal groups forming assemblies via synchronization based on experimentally obtained spike trains. The dynamics of such groups is reflected in the sequence of synchronized states, which we describe as a renewal dynamics. We furthermore introduce a rate function which is dependent on the internal network phase that quantifies the activity of neurons contributing to the observed spike train. This constitutes a hidden state model which is formally equivalent to a hidden Markov model, and all its parameters can be accurately determined from the experimental time series using the Baum-Welch algorithm. We apply our method to recordings from the cat visual cortex which exhibit oscillations and synchronizations. The parameters obtained for the hidden state model uncover characteristic properties of the system including synchronization, oscillation, switching, background activity and correlations. In applications involving multielectrode recordings, the extracted models quantify the extent of assembly formation and can be used for a temporally precise localization of system states underlying a specific spike train. Received: 30 March 1993/Accepted in revised form: 16 April 1994     

Nonlinear Dynamics of Emotion-Cognition Interaction: When Emotion Does not Destroy Cognition?

Emotion (i.e., spontaneous motivation and subsequent implementation of a behavior) and cognition (i.e., problem solving by information processing) are essential to how we, as humans, respond to changes in our environment. Recent studies in cognitive science suggest that emotion and cognition are subserved by different, although heavily integrated, neural systems. Understanding the time-varying relationship of emotion and cognition is a challenging goal with important implications for neuroscience. We formulate here the dynamical model of emotion-cognition interaction that is based on the following principles: (1) the temporal evolution of cognitive and emotion modes are captured by the incoming stimuli and competition within and among themselves (competition principle); (2) metastable states exist in the unified emotion-cognition phase space; and (3) the brain processes information with robust and reproducible transients through the sequence of metastable states. Such a model can take advantage of the often ignored temporal structure of the emotion-cognition interaction to provide a robust and generalizable method for understanding the relationship between brain activation and complex human behavior. The mathematical image of the robust and reproducible transient dynamics is a Stable Heteroclinic Sequence (SHS), and the Stable Heteroclinic Channels (SHCs). These have been hypothesized to be possible mechanisms that lead to the sequential transient behavior observed in networks. We investigate the modularity of SHCs, i.e., given a SHS and a SHC that is supported in one part of a network, we study conditions under which the SHC pertaining to the cognition will continue to function in the presence of interfering activity with other parts of the network, i.e., emotion.     

Floral morphogenesis: stochastic explorations of a gene network epigenetic landscape

In contrast to the classical view of development as a preprogrammed and deterministic process, recent studies have demonstrated that stochastic perturbations of highly non-linear systems may underlie the emergence and stability of biological patterns. Herein, we address the question of whether noise contributes to the generation of the stereotypical temporal pattern in gene expression during flower development. We modeled the regulatory network of organ identity genes in the Arabidopsis thaliana flower as a stochastic system. This network has previously been shown to converge to ten fixed-point attractors, each with gene expression arrays that characterize inflorescence cells and primordial cells of sepals, petals, stamens, and carpels. The network used is binary, and the logical rules that govern its dynamics are grounded in experimental evidence. We introduced different levels of uncertainty in the updating rules of the network. Interestingly, for a level of noise of around 0.5-10%, the system exhibited a sequence of transitions among attractors that mimics the sequence of gene activation configurations observed in real flowers. We also implemented the gene regulatory network as a continuous system using the Glass model of differential equations, that can be considered as a first approximation of kinetic-reaction equations, but which are not necessarily equivalent to the Boolean model. Interestingly, the Glass dynamics recover a temporal sequence of attractors, that is qualitatively similar, although not identical, to that obtained using the Boolean model. Thus, time ordering in the emergence of cell-fate patterns is not an artifact of synchronous updating in the Boolean model. Therefore, our model provides a novel explanation for the emergence and robustness of the ubiquitous temporal pattern of floral organ specification. It also constitutes a new approach to understanding morphogenesis, providing predictions on the population dynamics of cells with different genetic configurations during development.     

Software reliability prediction using recurrent neural network with Bayesian regularization

A recurrent neural network modeling approach for software reliability prediction with respect to cumulative failure time is proposed. Our proposed network structure has the capability of learning and recognizing the inherent internal temporal property of cumulative failure time sequence. Further, by adding a penalty term of sum of network connection weights, Bayesian regularization is applied to our network training scheme to improve the generalization capability and lower the susceptibility of overfitting. The performance of our proposed approach has been tested using four real-time control and flight dynamic application data sets. Numerical results show that our proposed approach is robust across different software projects, and has a better performance with respect to both goodness-of-fit and next-step-predictability compared to existing neural network models for failure time prediction.     

Markov models and network analysis reveal sex‐specific differences in the space‐use of a coastal apex predator

Understanding the links between external variables such as habitat and interactions with conspecifics and animal space‐use is fundamental to developing effective management measures. In the marine realm, automated acoustic tracking has become a widely used method for monitoring the movement of free‐ranging animals, yet researchers generally lack robust methods for analysing the resulting spatial‐usage data. In this study, acoustic tracking data from male and female broadnose sevengill sharks Notorynchus cepedianus, collected in a system of coastal embayments in southeast Tasmania were analyzed to examine sex‐specific differences in the sharks’ coastal space‐use and test novel methods for the analysis of acoustic telemetry data. Sex‐specific space‐use of the broadnose sevengill shark from acoustic telemetry data was analysed in two ways: The recently proposed spatial network analysis of between‐receiver movements was employed to identify sex‐specific space‐use patterns. To include the full breadth of temporal information held in the data, movements between receivers were furthermore considered as transitions between states of a Markov chain, with the resulting transition probability matrix allowing the ranking of the relative importance of different parts of the study area. Both spatial network and Markov chain analysis revealed sex‐specific preferences of different sites within the study area. The identification of priority areas differed for the methods, due to the fact that in contrast to network analysis, our Markov chain approach preserves the chronological sequence of detections and accounts for both residency periods and movements. In addition to adding to our knowledge of the ecology of a globally distributed apex predator, this study presents a promising new step towards condensing the vast amounts of information collected with acoustic tracking technology into straightforward results which are directly applicable to the management and conservation of any species that meet the assumptions of our model.     

A novel approach for short-term load forecasting using support vector machines

A support vector machine (SVM) modeling approach for short-term load forecasting is proposed. The SVM learning scheme is applied to the power load data, forcing the network to learn the inherent internal temporal property of power load sequence. We also study the performance when other related input variables such as temperature and humidity are considered. The performance of our proposed SVM modeling approach has been tested and compared with feed-forward neural network and cosine radial basis function neural network approaches. Numerical results show that the SVM approach yields better generalization capability and lower prediction error compared to those neural network approaches.     

Autologistic network model on binary data for disease progression study

This paper focuses on analysis of spatiotemporal binary data with absorbing states. The research was motivated by a clinical study on amyotrophic lateral sclerosis (ALS), a neurological disease marked by gradual loss of muscle strength over time in multiple body regions. We propose an autologistic regression model to capture complex spatial and temporal dependencies in muscle strength among different muscles. As it is not clear how the disease spreads from one muscle to another, it may not be reasonable to define a neighborhood structure based on spatial proximity. Relaxing the requirement for prespecification of spatial neighborhoods as in existing models, our method identifies an underlying network structure empirically to describe the pattern of spreading disease. The model also allows the network autoregressive effects to vary depending on the muscles’ previous status. Based on the joint distribution derived from this autologistic model, the joint transition probabilities of responses among locations can be estimated and the disease status can be predicted in the next time interval. Model parameters are estimated through maximization of penalized pseudo‐likelihood. Postmodel selection inference was conducted via a bias‐correction method, for which the asymptotic distributions were derived. Simulation studies were conducted to evaluate the performance of the proposed method. The method was applied to the analysis of muscle strength loss from the ALS clinical study.     

A Spike-Timing Pattern Based Neural Network Model for the Study of Memory Dynamics

It is well accepted that the brain''s computation relies on spatiotemporal activity of neural networks. In particular, there is growing evidence of the importance of continuously and precisely timed spiking activity. Therefore, it is important to characterize memory states in terms of spike-timing patterns that give both reliable memory of firing activities and precise memory of firing timings. The relationship between memory states and spike-timing patterns has been studied empirically with large-scale recording of neuron population in recent years. Here, by using a recurrent neural network model with dynamics at two time scales, we construct a dynamical memory network model which embeds both fast neural and synaptic variation and slow learning dynamics. A state vector is proposed to describe memory states in terms of spike-timing patterns of neural population, and a distance measure of state vector is defined to study several important phenomena of memory dynamics: partial memory recall, learning efficiency, learning with correlated stimuli. We show that the distance measure can capture the timing difference of memory states. In addition, we examine the influence of network topology on learning ability, and show that local connections can increase the network''s ability to embed more memory states. Together theses results suggest that the proposed system based on spike-timing patterns gives a productive model for the study of detailed learning and memory dynamics.     

Robustness elasticity in complex networks

Network robustness refers to a network's resilience to stress or damage. Given that most networks are inherently dynamic, with changing topology, loads, and operational states, their robustness is also likely subject to change. However, in most analyses of network structure, it is assumed that interaction among nodes has no effect on robustness. To investigate the hypothesis that network robustness is not sensitive or elastic to the level of interaction (or flow) among network nodes, this paper explores the impacts of network disruption, namely arc deletion, over a temporal sequence of observed nodal interactions for a large Internet backbone system. In particular, a mathematical programming approach is used to identify exact bounds on robustness to arc deletion for each epoch of nodal interaction. Elasticity of the identified bounds relative to the magnitude of arc deletion is assessed. Results indicate that system robustness can be highly elastic to spatial and temporal variations in nodal interactions within complex systems. Further, the presence of this elasticity provides evidence that a failure to account for nodal interaction can confound characterizations of complex networked systems.     

Likelihood analysis of phylogenetic networks using directed graphical models

A method for computing the likelihood of a set of sequences assuming a phylogenetic network as an evolutionary hypothesis is presented. The approach applies directed graphical models to sequence evolution on networks and is a natural generalization of earlier work by Felsenstein on evolutionary trees, including it as a special case. The likelihood computation involves several steps. First, the phylogenetic network is rooted to form a directed acyclic graph (DAG). Then, applying standard models for nucleotide/amino acid substitution, the DAG is converted into a Bayesian network from which the joint probability distribution involving all nodes of the network can be directly read. The joint probability is explicitly dependent on branch lengths and on recombination parameters (prior probability of a parent sequence). The likelihood of the data assuming no knowledge of hidden nodes is obtained by marginalization, i.e., by summing over all combinations of unknown states. As the number of terms increases exponentially with the number of hidden nodes, a Markov chain Monte Carlo procedure (Gibbs sampling) is used to accurately approximate the likelihood by summing over the most important states only. Investigating a human T-cell lymphotropic virus (HTLV) data set and optimizing both branch lengths and recombination parameters, we find that the likelihood of a corresponding phylogenetic network outperforms a set of competing evolutionary trees. In general, except for the case of a tree, the likelihood of a network will be dependent on the choice of the root, even if a reversible model of substitution is applied. Thus, the method also provides a way in which to root a phylogenetic network by choosing a node that produces a most likely network.     

Chaos and synchrony in a model of a hypercolumn in visual cortex

Neurons in cortical slices emit spikes or bursts of spikes regularly in response to a suprathreshold current injection. This behavior is in marked contrast to the behavior of cortical neurons in vivo, whose response to electrical or sensory input displays a strong degree of irregularity. Correlation measurements show a significant degree of synchrony in the temporal fluctuations of neuronal activities in cortex. We explore the hypothesis that these phenomena are the result of the synchronized chaos generated by the deterministic dynamics of local cortical networks. A model of a hypercolumn in the visual cortex is studied. It consists of two populations of neurons, one inhibitory and one excitatory. The dynamics of the neurons is based on a Hodgkin-Huxley type model of excitable voltage-clamped cells with several cellular and synaptic conductances. A slow potassium current is included in the dynamics of the excitatory population to reproduce the observed adaptation of the spike trains emitted by these neurons. The pattern of connectivity has a spatial structure which is correlated with the internal organization of hypercolumns in orientation columns. Numerical simulations of the model show that in an appropriate parameter range, the network settles in a synchronous chaotic state, characterized by a strong temporal variability of the neural activity which is correlated across the hypercolumn. Strong inhibitory feedback is essential for the stabilization of this state. These results show that the cooperative dynamics of large neuronal networks are capable of generating variability and synchrony similar to those observed in cortex. Auto-correlation and cross-correlation functions of neuronal spike trains are computed, and their temporal and spatial features are analyzed. In other parameter regimes, the network exhibits two additional states: synchronized oscillations and an asynchronous state. We use our model to study cortical mechanisms for orientation selectivity. It is shown that in a suitable parameter regime, when the input is not oriented, the network has a continuum of states, each representing an inhomogeneous population activity which is peaked at one of the orientation columns. As a result, when a weakly oriented input stimulates the network, it yields a sharp orientation tuning. The properties of the network in this regime, including the appearance of virtual rotations and broad stimulus-dependent cross-correlations, are investigated. The results agree with the predictions of the mean field theory which was previously derived for a simplified model of stochastic, two-state neurons. The relation between the results of the model and experiments in visual cortex are discussed.     

Stochastic E2F Activation and Reconciliation of Phenomenological Cell-Cycle Models

The transition of the mammalian cell from quiescence to proliferation is a highly variable process. Over the last four decades, two lines of apparently contradictory, phenomenological models have been proposed to account for such temporal variability. These include various forms of the transition probability (TP) model and the growth control (GC) model, which lack mechanistic details. The GC model was further proposed as an alternative explanation for the concept of the restriction point, which we recently demonstrated as being controlled by a bistable Rb-E2F switch. Here, through a combination of modeling and experiments, we show that these different lines of models in essence reflect different aspects of stochastic dynamics in cell cycle entry. In particular, we show that the variable activation of E2F can be described by stochastic activation of the bistable Rb-E2F switch, which in turn may account for the temporal variability in cell cycle entry. Moreover, we show that temporal dynamics of E2F activation can be recast into the frameworks of both the TP model and the GC model via parameter mapping. This mapping suggests that the two lines of phenomenological models can be reconciled through the stochastic dynamics of the Rb-E2F switch. It also suggests a potential utility of the TP or GC models in defining concise, quantitative phenotypes of cell physiology. This may have implications in classifying cell types or states.