Span: spike pattern association neuron for learning spatio-temporal spike patterns

Spiking Neural Networks (SNN) were shown to be suitable tools for the processing of spatio-temporal information. However, due to their inherent complexity, the formulation of efficient supervised learning algorithms for SNN is difficult and remains an important problem in the research area. This article presents SPAN - a spiking neuron that is able to learn associations of arbitrary spike trains in a supervised fashion allowing the processing of spatio-temporal information encoded in the precise timing of spikes. The idea of the proposed algorithm is to transform spike trains during the learning phase into analog signals so that common mathematical operations can be performed on them. Using this conversion, it is possible to apply the well-known Widrow-Hoff rule directly to the transformed spike trains in order to adjust the synaptic weights and to achieve a desired input/output spike behavior of the neuron. In the presented experimental analysis, the proposed learning algorithm is evaluated regarding its learning capabilities, its memory capacity, its robustness to noisy stimuli and its classification performance. Differences and similarities of SPAN regarding two related algorithms, ReSuMe and Chronotron, are discussed.     

The gradient clusteron: A model neuron that learns to solve classification tasks via dendritic nonlinearities,structural plasticity,and gradient descent

Synaptic clustering on neuronal dendrites has been hypothesized to play an important role in implementing pattern recognition. Neighboring synapses on a dendritic branch can interact in a synergistic, cooperative manner via nonlinear voltage-dependent mechanisms, such as NMDA receptors. Inspired by the NMDA receptor, the single-branch clusteron learning algorithm takes advantage of location-dependent multiplicative nonlinearities to solve classification tasks by randomly shuffling the locations of “under-performing” synapses on a model dendrite during learning (“structural plasticity”), eventually resulting in synapses with correlated activity being placed next to each other on the dendrite. We propose an alternative model, the gradient clusteron, or G-clusteron, which uses an analytically-derived gradient descent rule where synapses are "attracted to" or "repelled from" each other in an input- and location-dependent manner. We demonstrate the classification ability of this algorithm by testing it on the MNIST handwritten digit dataset and show that, when using a softmax activation function, the accuracy of the G-clusteron on the all-versus-all MNIST task (~85%) approaches that of logistic regression (~93%). In addition to the location update rule, we also derive a learning rule for the synaptic weights of the G-clusteron (“functional plasticity”) and show that a G-clusteron that utilizes the weight update rule can achieve ~89% accuracy on the MNIST task. We also show that a G-clusteron with both the weight and location update rules can learn to solve the XOR problem from arbitrary initial conditions.     

Continuous functions determined by spike trains of a neuron subject to stimulation

several ways of estimating a continuous function from the spike train output of a neuron subjected to repeated stimuli are compared: (i) the probability of firing function estimated by a PST-histogram (ii) the rate of discharge function estimated by a frequencygram (Bessou et al. 1968) and (iii) the interspike-interval function which is introduced in this paper. For a special class of neuronal responses, called deterministic, these functions may be expressed in terms of each other. It is shown that the current clamped Hodgkin-Huxley model of an action potential encoding membrane (Hodgkin and Huxley 1952) is able to generate such deterministic responses. As an experimental example, a deterministic response of a primary muscle spindle afferent is used to demonstrate the estimation of the functions. Interpretability and numerical estimatability of these spike train describing functions are discussed for deterministic neuronal responses.     

Dynamical estimation of neuron and network properties III: network analysis using neuron spike times

Estimating the behavior of a network of neurons requires accurate models of the individual neurons along with accurate characterizations of the connections among them. Whereas for a single cell, measurements of the intracellular voltage are technically feasible and sufficient to characterize a useful model of its behavior, making sufficient numbers of simultaneous intracellular measurements to characterize even small networks is infeasible. This paper builds on prior work on single neurons to explore whether knowledge of the time of spiking of neurons in a network, once the nodes (neurons) have been characterized biophysically, can provide enough information to usefully constrain the functional architecture of the network: the existence of synaptic links among neurons and their strength. Using standardized voltage and synaptic gating variable waveforms associated with a spike, we demonstrate that the functional architecture of a small network of model neurons can be established.     

Determination of the fraction of active inputs required by a neuron to fire

What fraction of the inputs to a neuron in the primary visual cortex (V1) need to be active for that neuron to reach its firing threshold? The paper describes a numerical method for estimating the selectivity of visual neurons, in terms of the required fraction of active excitatory inputs, from standard data produced by intracellular electro-physiological recordings. The method also provides an estimate of the relative strength of the feedforward inhibition in a push-pull model of the inputs to V1 simple cells. The method is tested on two V1 cells described in Carandini and Ferster [Carandini, M., Ferster, D., 2000. Membrane potential and firing rate in cat primary visual cortex. J. Neurosci. 20, 470-484]. The results indicate that the maximum strength of feedforward inhibition is around 30% of the maximum strength of feedforward excitation. The two V1 neurons investigated fire if more than around 40% of their excitatory LGN inputs are active.     

Multiple spike initiation zones in a neuron implicated in learning in the leech: a computational model

Sensitization of the defensive shortening reflex in the leech has been linked to a segmentally repeated tri-synaptic positive feedback loop. Serotonin from the R-cell enhances S-cell excitability, S-cell impulses cross an electrical synapse into the C-interneuron, and the C-interneuron excites the R-cell via a glutamatergic synapse. The C-interneuron has two unusual characteristics. First, impulses take longer to propagate from the S soma to the C soma than in the reverse direction. Second, impulses recorded from the electrically unexcitable C soma vary in amplitude when extracellular divalent cation concentrations are elevated, with smaller impulses failing to induce synaptic potentials in the R-cell. A compartmental, computational model was developed to test the sufficiency of multiple, independent spike initiation zones in the C-interneuron to explain these observations. The model displays asymmetric delays in impulse propagation across the S–C electrical synapse and graded impulse amplitudes in the C-interneuron in simulated high divalent cation concentrations.     

Herbivore host plant selection: whitefly learns to avoid host plants that harbour predators of her offspring

Evidence is accumulating that herbivorous arthropods do not simply select host plants based on their quality, but also on the predation risk associated with different host plants. It has been suggested that herbivores exclude plant species with high predation risk from their host range. This assumes a constant, predictable predation risk as well as a rather static behaviour on the part of the herbivore; plants are ignored irrespective of the actual predation risk. We show that adult females of a small herbivore, the whitefly Bemisia tabaci, can learn to avoid plants with predatory mites that attack only juvenile whiteflies, while they accept host plants of the same species without predators. Predatory mites disperse more slowly than whiteflies; they cannot fly and walk from plant to plant. Hence, by avoiding plants with predators, the whiteflies create a temporary refuge for their offspring. We suggest that the experience of arthropod herbivores with risks associated with host plants plays an important role in their host plant selection.     

The BMP signaling gradient patterns dorsoventral tissues in a temporally progressive manner along the anteroposterior axis

Patterning of the vertebrate anteroposterior (AP) axis proceeds temporally from anterior to posterior. How dorsoventral (DV) axial patterning relates to AP temporal patterning is unknown. We examined the temporal activity of BMP signaling in patterning ventrolateral cell fates along the AP axis, using transgenes that rapidly turn "off" or "on" BMP signaling. We show that BMP signaling patterns rostral DV cell fates at the onset of gastrulation, whereas progressively more caudal DV cell fates are patterned at progressively later intervals during gastrulation. Increased BMP signal duration is not required to pattern more caudal DV cell fates; rather, distinct temporal intervals of signaling are required. This progressive action is regulated downstream of, or in parallel to, BMP signal transduction at the level of Smad1/5 phosphorylation. We propose that a temporal cue regulates a cell's competence to respond to BMP signaling, allowing the acquisition of a cell's DV and AP identity simultaneously.     

NMDA receptor-dependent long-term potentiation comprises a family of temporally overlapping forms of synaptic plasticity that are induced by different patterns of stimulation

N-methyl-d-aspartate receptor (NMDAR)-dependent long-term potentiation (LTP) is extensively studied since it is believed to use the same molecular mechanisms that are required for many forms of learning and memory. Unfortunately, many controversies exist, not least the seemingly simple issue concerning the locus of expression of LTP. Here, we review our recent work and some of the extensive literature on this topic and present new data that collectively suggest that LTP can be explained, during its first few hours, by the coexistence of at least three mechanistically distinct processes that are all triggered by the synaptic activation of NMDARs.     

Characterization and precise mapping of a QTL increasing spike number with pleiotropic effects in wheat

Tiller number (TN) and spike number per plant (SN) are key components of grain yield and/or biomass in wheat. In this study, an introgression line 05210, developed by introgression of chromosomal segments from a synthetic exotic wheat Am3 into an elite cultivar Laizhou953, showed a significantly increased TN and SN, but shorter spike length (SL) and fewer grain number per spike (GNS) than Laizhou953. To investigate the quantitative trait locus (QTL) responsible for these variations, the introgressed segments in 05210 were screened by SSR markers and one follow-up segregation population was developed from the cross 05210/Laizhou953. The population showed 3:1 segregation ratios for SN, SL and GNS, indicating that QTLs for these traits have been dissected into single Mendelian factors. Bulked segregation analysis showed that the markers located on the 4B introgressed segment were polymorphic between the two bulks. Therefore, they were further analyzed in the F₂ population to construct a linkage map. Three new QTLs, QSn.sdau-4B, QSl.sdau-4B and QGns.sdau-4B, were detected for SN, SL and GNS, respectively, which explained a large portion of the phenotypic variation (30.1–67.6%) for these traits with overlapping peaks. Correlation analysis and multiple-trait, multiple-interval mapping (MMIM) suggested pleiotropic effects of the QTL on SN, SL and GNS. Therefore, the QTL was designated as QSn.sdau-4B. By a progeny test based on F₃ families using SN, the QTL was mapped as a Mendelian factor to the proximal region of 4BL. It is a key QTL responsible for variation in spike number and size, which had not been reported previously. Thus, it is an important QTL for wheat to achieve high and stable biomass and grain yield. Dissection and mapping of this QTL as a Mendelian factor laid a solid foundation for map-based cloning of grain yield-related QTLs in wheat.     

Patch to landscape patterns in post fire recruitment of a serotinous conifer

Obligate seeding species are highly specialized to fire disturbance and many conifers such as cypress, which are adapted to high intensity stand-replacing fires, have canopy seed banks stored in serotinous cones. Resilience of these trees to fire disturbance is a function of disturbance frequency and one focus of this study was to determine the effect of patch age on postfire recruitment. A second focus was to determine the extent to which fire induced a landscape level change in the location of the forest boundary. Prior to a fire in 1994, a large Cupressus sargentii forest was a mosaic landscape of different aged patches of nearly pure cypress bordered by chaparral. Patches less than 60 years of age were relatively dense with roughly one tree every 1–2 m² but older patches had thinned to one tree every 3–15 m². Older trees had substantially greater canopy cone crops but the stand level seed bank size was not significantly correlated with stand age. Fire-dependent obligate seeding species are sensitive to fire return interval because of potential changes in the size of seed banks – facing both a potential `immaturity risk' and a `senescence risk'. At our site, C. sargentii regeneration was substantial in stands as young as 20 years, suggesting that fire return interval would need to be shorter than this to pose any significant risk. Reduced seedling recruitment in stands nearly 100 years of age may indicate risk from senescence is greater, however, even the lowest density seedling recruitment was many times greater than the density of mature forests – thus this cypress would appear to be resilient to a wide range of fire return intervals. Changes in landscape patterning of forest and chaparral are unlikely except after fire. Factors that inhibit tree establishment within the shrubland, as well as factors that affect shrub establishment within the forest border likely affect the `permeability' of this ecotone. After the 1994 fire this boundary appeared to be stable in that cypress recruited best within the shadow of burned canopies and cypress were weak invaders of adjacent shrublands.     

The encoding of temporally irregular and regular visual patterns in the human brain

In the work reported here, we set out to study the neural systems that detect predictable temporal patterns and departures from them. We used functional magnetic resonance imaging (fMRI) to locate activity in the brains of subjects when they viewed temporally regular and irregular patterns produced by letters, numbers, colors and luminance. Activity induced by irregular sequences was located within dorsolateral prefrontal cortex, including an area that was responsive to irregular patterns regardless of the type of visual stimuli producing them. Conversely, temporally regular arrangements resulted in activity in the right frontal lobe (medial frontal gyrus), in the left orbito-frontal cortex and in the left pallidum. The results show that there is an abstractive system in the brain for detecting temporal irregularity, regardless of the source producing it.     

A hidden Markov model approach to neuron firing patterns.

Analysis and characterization of neuronal discharge patterns are of interest to neurophysiologists and neuropharmacologists. In this paper we present a hidden Markov model approach to modeling single neuron electrical activity. Basically the model assumes that each interspike interval corresponds to one of several possible states of the neuron. Fitting the model to experimental series of interspike intervals by maximum likelihood allows estimation of the number of possible underlying neuron states, the probability density functions of interspike intervals corresponding to each state, and the transition probabilities between states. We present an application to the analysis of recordings of a locus coeruleus neuron under three pharmacological conditions. The model distinguishes two states during halothane anesthesia and during recovery from halothane anesthesia, and four states after administration of clonidine. The transition probabilities yield additional insights into the mechanisms of neuron firing.     

The Decapentaplegic morphogen gradient: a precise definition

Two key processes are in the basis of morphogenesis: the spatial allocation of cell types in fields of na?ve cells and the regulation of growth. Both are controlled by morphogens, which activate target genes in the growing tissue in a concentration-dependent manner. Thus the morphogen model is an intrinsically quantitative concept. However, quantitative studies were performed only in recent years on two morphogens: Bicoid and Decapentaplegic. This review covers quantitative aspects of the formation and precision of the Decapentaplegic morphogen gradient. The morphogen gradient concept is transitioning from a soft definition to a precise idea of what the gradient could really do.     

The boundary cap: a source of neural crest stem cells that generate multiple sensory neuron subtypes

The boundary cap (BC) is a transient neural crest-derived group of cells located at the dorsal root entry zone (DREZ) that have been shown to differentiate into sensory neurons and glia in vivo. We find that when placed in culture, BC cells self-renew, show multipotency in clonal cultures and express neural crest stem cell (NCSCs) markers. Unlike sciatic nerve NCSCs, the BC-NCSC (bNCSCs) generates sensory neurons upon differentiation. The bNCSCs constitute a common source of cells for functionally diverse types of neurons, as a single bNCSC can give rise to several types of nociceptive and thermoreceptive sensory neurons. Our data suggests that BC cells comprise a source of multipotent sensory specified stem cells that persist throughout embryogenesis.     

Antibodies that recognize astrotactin block granule neuron binding to astroglia

To provide a rapid, specific assay for receptor systems involved in the binding of cerebellar granule neurons to astroglia, granule cells, purified from early postnatal mice, or from E15-E16 chicks, were radiolabeled with [35S]methionine and plasma membranes were prepared. The kinetics of binding of radiolabeled material to primary mouse or chick glia or to the mouse G26-24 astrocytoma cell line was measured in the presence or absence of antibodies against astrotactin, neural cell adhesion molecules, cadherins, or integrins. Addition of Fab fragments of astrotactin antibodies reduced the amount of granule cell membrane binding to astroglia by 70%. In contrast, Fab fragments of antibodies against the neural adhesion molecules N-CAM, L1, and N-cadherin and against integrin did not reduce the level of granule cell membrane binding to astroglia. Combinations of antibodies against N-CAM, L1, N-cadherin, and integrin also did not impair neuron binding to glia.