Anatomical funneling, sparse connectivity and redundancy reduction in the neural networks of the basal ganglia

Sparse non-convex regularization based explainable DBN in the analysis of brain abnormalities in schizophrenia

Deep belief networks have been widely used in medical image analysis. However, the high-dimensional but small-sample-size characteristic of medical image data makes the model prone to dimensional disaster and overfitting. Meanwhile, the traditional DBN is driven by performance and ignores the explainability which is important for medical image analysis. In this paper, a sparse non-convex based explainable deep belief network is proposed by combining DBN with non-convex sparsity learning. For sparsity, the non-convex regularization and Kullback-Leibler divergence penalty are embedded into DBN to obtain the sparse connection and sparse response representation of the network. It effectively reduces the complexity of the model and improves the generalization ability of the model. Considering explainability, the crucial features for decision-making are selected through the feature back-selection based on the row norm of each layer's weight after network training. We apply the model to schizophrenia data and demonstrate it achieves the best performance among several typical feature selection models. It reveals 28 functional connections highly correlated with schizophrenia, which provides an effective foundation for the treatment and prevention of schizophrenia and methodological assurance for similar brain disorders.

Structure Learning for Deep Neural Networks Based on Multiobjective Optimization

This paper focuses on the connecting structure of deep neural networks and proposes a layerwise structure learning method based on multiobjective optimization. A model with better generalization can be obtained by reducing the connecting parameters in deep networks. The aim is to find the optimal structure with high representation ability and better generalization for each layer. Then, the visible data are modeled with respect to structure based on the products of experts. In order to mitigate the difficulty of estimating the denominator in PoE, the denominator is simplified and taken as another objective, i.e., the connecting sparsity. Moreover, for the consideration of the contradictory nature between the representation ability and the network connecting sparsity, the multiobjective model is established. An improved multiobjective evolutionary algorithm is used to solve this model. Two tricks are designed to decrease the computational cost according to the properties of input data. The experiments on single-layer level, hierarchical level, and application level demonstrate the effectiveness of the proposed algorithm, and the learned structures can improve the performance of deep neural networks.

Dysregulation of striatal projection neurons in Parkinson's disease

The loss of nigrostriatal dopamine (DA) is the primary cause of motor dysfunction in Parkinson's disease (PD), but the underlying striatal mechanisms remain unclear. In spite of abundant literature portraying structural, biochemical and plasticity changes of striatal projection neurons (SPNs), in the past there has been a data vacuum from the natural human disease and its close model in non-human primates. Recently, single-cell recordings in advanced parkinsonian primates have generated new insights into the altered function of SPNs. Currently, there are also human data that provide direct evidence of profoundly dysregulated SPN activity in PD. Here, we review primate recordings that are impacting our understanding of the striatal dysfunction after DA loss, particularly through the analysis of physiologic correlates of parkinsonian motor behaviors. In contrast to recordings in rodents, data obtained in primates and patients demonstrate similar major abnormalities of the spontaneous SPN firing in the alert parkinsonian state. Furthermore, these studies also show altered SPN responses to DA replacement in the advanced parkinsonian state. Clearly, there is yet much to learn about the striatal discharges in PD, but studies using primate models are contributing unique information to advance our understanding of pathophysiologic mechanisms.

Sparse pallidal connections shape synchrony in a network model of the basal ganglia

Neural synchrony in the basal ganglia, especially in the beta frequency band (13-30 Hz), is a hallmark of Parkinson's disease and considered as antikinetic. In contrast, the healthy basal ganglia show low levels of synchrony. It is currently unknown where synchrony and oscillations arise in the parkinsonian brain and how they are transmitted through the basal ganglia, as well as what makes them dependent on dopamine. The external part of the globus pallidus has recently been identified as a hub nucleus in the basal ganglia, possessing intrinsic inhibitory connections and possibly also gap junctions. In this study, we show that in a conductance-based network model of the basal ganglia, the combination of sparse, high-conductance inhibitory synapses and sparse, low-conductance gap junctions in the external part of the globus pallidus could effectively desynchronize the whole network. However, when gap junction coupling became strong enough, the effect was impeded and activity synchronized. In particular, sustained periods of beta coherence occurred between some neuron pairs. As gap junctions can change their conductance with the dopamine level, we suggest pallidal gap junction coupling as a mechanism contributing to the development of beta synchrony in the parkinsonian basal ganglia.

A Multiobjective Sparse Feature Learning Model for Deep Neural Networks

Hierarchical deep neural networks are currently popular learning models for imitating the hierarchical architecture of human brain. Single-layer feature extractors are the bricks to build deep networks. Sparse feature learning models are popular models that can learn useful representations. But most of those models need a user-defined constant to control the sparsity of representations. In this paper, we propose a multiobjective sparse feature learning model based on the autoencoder. The parameters of the model are learnt by optimizing two objectives, reconstruction error and the sparsity of hidden units simultaneously to find a reasonable compromise between them automatically. We design a multiobjective induced learning procedure for this model based on a multiobjective evolutionary algorithm. In the experiments, we demonstrate that the learning procedure is effective, and the proposed multiobjective model can learn useful sparse features.

Properties of VIP+ synapses in the suprachiasmatic nucleus highlight their role in circadian rhythm

Circadian rhythms coordinate cyclical behavioral and physiological changes in most organisms. In humans, this biological clock is located within the suprachiasmatic nucleus (SCN) of the hypothalamus and consists of a heterogeneous neuron population characterized by their enriched expression of various neuropeptides. As highlighted here, Fan et al. (J Neurosci 35: 1905-1029, 2015) developed an elegant experimental system to investigate the synaptic properties of vasoactive intestinal peptide (VIP)-expressing neurons between day and night, and further delineate their broader architecture and function within the SCN.

Neuronal circuits underlying persistent representations despite time varying activity

BACKGROUND

Our brains are capable of remarkably stable stimulus representations despite time-varying neural activity. For instance, during delay periods in working memory tasks, while stimuli are represented in working memory, neurons in the prefrontal cortex, thought to support the memory representation, exhibit time-varying neuronal activity. Since neuronal activity encodes the stimulus, its time-varying dynamics appears to be paradoxical and incompatible with stable network stimulus representations. Indeed, this finding raises a fundamental question: can stable representations only be encoded with stable neural activity, or, its corollary, is every change in activity a sign of change in stimulus representation?

RESULTS

Here we explain how different time-varying representations offered by individual neurons can be woven together to form a coherent, time-invariant, representation. Motivated by two ubiquitous features of the neocortex-redundancy of neural representation and sparse intracortical connections-we derive a network architecture that resolves the apparent contradiction between representation stability and changing neural activity. Unexpectedly, this network architecture exhibits many structural properties that have been measured in cortical sensory areas. In particular, we can account for few-neuron motifs, synapse weight distribution, and the relations between neuronal functional properties and connection probability.

CONCLUSIONS

We show that the intuition regarding network stimulus representation, typically derived from considering single neurons, may be misleading and that time-varying activity of distributed representation in cortical circuits does not necessarily imply that the network explicitly encodes time-varying properties.

How do the basal ganglia contribute to categorization? Their roles in generalization, response selection, and learning via feedback

This article examines how independent corticostriatal loops linking basal ganglia with cerebral cortex contribute to visual categorization. The first aspect of categorization discussed is the role of the visual corticostriatal loop, which connects the visual cortex and the body/tail of the caudate, in mapping visual stimuli to categories, including evaluating the degree to which this loop may generalize across individual category members. The second aspect of categorization discussed is the selection of appropriate actions or behaviors on the basis of category membership, and the role of the visual corticostriatal loop output and the motor corticostriatal loop, which connects motor planning areas with the putamen, in action selection. The third aspect of categorization discussed is how categories are learned with the aid of feedback linked dopaminergic projections to the basal ganglia. These projections underlie corticostriatal synaptic plasticity across the basal ganglia, and also serve as input to the executive and motivational corticostriatal loops that play a role in strategic use of feedback.

A physiologically plausible model of action selection and oscillatory activity in the basal ganglia

The basal ganglia (BG) have long been implicated in both motor function and dysfunction. It has been proposed that the BG form a centralized action selection circuit, resolving conflict between multiple neural systems competing for access to the final common motor pathway. We present a new spiking neuron model of the BG circuitry to test this proposal, incorporating all major features and many physiologically plausible details. We include the following: effects of dopamine in the subthalamic nucleus (STN) and globus pallidus (GP), transmission delays between neurons, and specific distributions of synaptic inputs over dendrites. All main parameters were derived from experimental studies. We find that the BG circuitry supports motor program selection and switching, which deteriorates under dopamine-depleted and dopamine-excessive conditions in a manner consistent with some pathologies associated with those dopamine states. We also validated the model against data describing oscillatory properties of BG. We find that the same model displayed detailed features of both gamma-band (30-80 Hz) and slow (approximately 1 Hz) oscillatory phenomena reported by Brown et al. (2002) and Magill et al. (2001), respectively. Only the parameters required to mimic experimental conditions (e.g., anesthetic) or manipulations (e.g., lesions) were changed. From the results, we derive the following novel predictions about the STN-GP feedback loop: (1) the loop is functionally decoupled by tonic dopamine under normal conditions and recoupled by dopamine depletion; (2) the loop does not show pacemaking activity under normal conditions in vivo (but does after combined dopamine depletion and cortical lesion); (3) the loop has a resonant frequency in the gamma-band.