Self-organized phase transitions in neural networks as a neural mechanism of information processing

Transitions between dynamically stable activity patterns imposed on an associative neural network are shown to be induced by self-organized infinitesimal changes in synaptic connection strength and to be a kind of phase transition. A key event for the neural process of information processing in a population coding scheme is transition between the activity patterns encoding usual entities. We propose that the infinitesimal and short-term synaptic changes based on the Hebbian learning rule are the driving force for the transition. The phase transition between the following two dynamical stable states is studied in detail, the state where the firing pattern is changed temporally so as to itinerate among several patterns and the state where the firing pattern is fixed to one of several patterns. The phase transition from the pattern itinerant state to a pattern fixed state may be induced by the Hebbian learning process under a weak input relevant to the fixed pattern. The reverse transition may be induced by the Hebbian unlearning process without input. The former transition is considered as recognition of the input stimulus, while the latter is considered as clearing of the used input data to get ready for new input. To ensure that information processing based on the phase transition can be made by the infinitesimal and short-term synaptic changes, it is absolutely necessary that the network always stays near the critical state corresponding to the phase transition point.     

Making time count: Functional evidence for temporal coding of taste sensation.

Although the temporal characteristics of neural responses have been proposed as a mechanism for sensory neural coding, there has been little evidence thus far that this type of information is actually used by the nervous system. Here the authors show that patterned electrical pulses trains that mimic the response to the taste of quinine can produce a bitterlike sensation when delivered to the nucleus tractus solitarius of behaving rats. Following conditioned aversion training using either "quinine simulation" patterns of electrical stimulation or natural quinine (0.1 mM) as a conditioned stimulus, rats specifically generalized the aversion to 2 bitter tastants: quinine and urea. Randomization of the quinine simulation patterns resulted in generalization patterns that resembled those to a perithreshold concentration (0.01 mM) of quinine. These data provide strong evidence that the temporal pattern of brainstem activity may convey information about taste quality and underscore the functional significance of temporal coding. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Fast sigmoidal networks via spiking neurons

We show that networks of relatively realistic mathematical models for biological neurons in principle can simulate arbitrary feedforward sigmoidal neural nets in a way that has previously not been considered. This new approach is based on temporal coding by single spikes (respectively by the timing of synchronous firing in pools of neurons) rather than on the traditional interpretation of analog variables in terms of firing rates. The resulting new simulation is substantially faster and hence more consistent with experimental results about the maximal speed of information processing in cortical neural systems. As a consequence we can show that networks of noisy spiking neurons are "universal approximators" in the sense that they can approximate with regard to temporal coding any given continuous function of several variables. This result holds for a fairly large class of schemes for coding analog variables by firing times of spiking neurons. This new proposal for the possible organization of computations in networks of spiking neurons systems has some interesting consequences for the type of learning rules that would be needed to explain the self-organization of such networks. Finally, the fast and noise-robust implementation of sigmoidal neural nets by temporal coding points to possible new ways of implementing feedforward and recurrent sigmoidal neural nets with pulse stream VLSI.     

Engrams, memory storage, and mnemonic coding.

Among the basic questions concerning memory is the problem of "how the human brain records, stores, and transmits information… . I believe that acquired engrams are laid down like genetic codes, are transmitted as genetic information and revived and assembled into memories by some matching method. The matching may be either chemical or electrical and at some levels of neural organization of the sort we identify with frequency radio tuning or electrical resonance… . [If] we conceive of learning as a linking of sequential responses or facts, it is not the facts but the links between them which are altered by the learning process." Brain function resembles genetic replication and resonance phenomena rather than devices currently employed in automata. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Multitasking in the olfactory system: context-dependent responses to odors reveal dual GABA-regulated coding mechanisms in single olfactory projection neurons

Studies of olfaction have focused mainly on neural processing of information about the chemistry of odors, but olfactory stimuli have other properties that also affect central responses and thus influence behavior. In moths, continuous and intermittent stimulation with the same odor evokes two distinct flight behaviors, but the neural basis of this differential response is unknown. Here we show that certain projection neurons (PNs) in the primary olfactory center in the brain give context-dependent responses to a specific odor blend, and these responses are shaped in several ways by a bicuculline-sensitive GABA receptor. Pharmacological dissection of PN responses reveals that bicuculline blocks GABAA-type receptors/chloride channels in PNs, and that these receptors play a critical role in shaping the responses of these glomerular output neurons. The firing patterns of PNs are not odor-specific but are strongly modulated by the temporal pattern of the odor stimulus. Brief repetitive odor pulses evoke fast inhibitory potentials, followed by discrete bursts of action potentials that are phase-locked to the pulses. In contrast, the response to a single prolonged stimulus with the same odor is a series of slow oscillations underlying irregular firing. Bicuculline disrupts the timing of both types of responses, suggesting that GABAA-like receptors underlie both coding mechanisms. These results suggest that glomerular output neurons could use more than one coding scheme to represent a single olfactory stimulus. Moreover, these context-dependent odor responses encode information about both the chemical composition and the temporal pattern of the odor signal. Together with behavioral evidence, these findings suggest that context-dependent odor responses evoke different perceptions in the brain that provide the animal with important information about the spatiotemporal variations that occur in natural odor plumes.     

Deciphering a neural code for vision

Deciphering the information that eyes, ears, and other sensory organs transmit to the brain is important for understanding the neural basis of behavior. Recordings from single sensory nerve cells have yielded useful insights, but single neurons generally do not mediate behavior; networks of neurons do. Monitoring the activity of all cells in a neural network of a behaving animal, however, is not yet possible. Taking an alternative approach, we used a realistic cell-based model to compute the ensemble of neural activity generated by one sensory organ, the lateral eye of the horseshoe crab, Limulus polyphemus. We studied how the neural network of this eye encodes natural scenes by presenting to the model movies recorded with a video camera mounted above the eye of an animal that was exploring its underwater habitat. Model predictions were confirmed by simultaneously recording responses from single optic nerve fibers of the same animal. We report here that the eye transmits to the brain robust "neural images" of objects having the size, contrast, and motion of potential mates. The neural code for such objects is not found in ambiguous messages of individual optic nerve fibers but rather in patterns of coherent activity that extend over small ensembles of nerve fibers and are bound together by stimulus motion. Integrative properties of neurons in the first synaptic layer of the brain appear well suited to detecting the patterns of coherent activity. Neural coding by this relatively simple eye helps explain how horseshoe crabs find mates and may lead to a better understanding of how more complex sensory organs process information.     

The search for the engram.

Reviews literature aimed at studying the nature of the "engram," the set of physical processes and changes in the brain which forms the basis of learning. The general research strategies currently employed are (a) model biological systems and (b) more or less intact behaving animals. Use of the model systems approach has supported the view that habituation reflects some form of synaptic depression while sensitization reflects some form of superimposed or heterosynaptic facilitation. Advances have also been made in the analysis of neural control of behavioral responses and response sequences in invertebrates. Study in the hindlimb flexion response of vertebrates has contributed to the formation of a simplified neuronal model of classical conditioning. Although simplified models of learning provide information about the nature of the basic synaptic processes involved, study of the mammalian CNS is necessary to understand the neural processes in human learning and memory. Neuronal activity in the hippocampus, a brain structure believed to be involved in learning, has suggested that the first process in the formation of the engram may occur in the hippocampus. Because the hippocampal response develops early in training and is large and reliable, localization of the initial formation of this learning-dependent brain response is feasible. Given the parallel processing in the brain, there are brain systems yet to be studied which may be involved in the formation of the engram. (4 p ref) (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Posttraining shifts in the overshadowing stimulus–unconditioned stimulus interval alleviates the overshadowing deficit.

Two conditioned lick suppression experiments explored the effects on overshadowing of a posttraining change in the temporal relationship between the overshadowing conditioned stimulus (CS) and the unconditioned stimulus (US). Rats received either trace (Experiment 1) or delay (Experiment 2) overshadowing training. Then pairings of the overshadowing CS and US were given with either a trace or delay temporal relationship. Overshadowing was alleviated by shifting the overshadowing CS–US temporal relationship so that it no longer matched the overshadowed CS–US temporal relationship. These outcomes are explicable in terms of an integration of the comparator hypothesis, which states that cue competition effects (e.g., overshadowing) will be maximal when the information potentially conveyed by competing CSs is equivalent, and the temporal coding hypothesis, which states that CS–US intervals are part of the information encoded during conditioning. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

The expression of the mouse Zic1, Zic2, and Zic3 gene suggests an essential role for Zic genes in body pattern formation

We examined the expression of Zic1, Zic2, and Zic3 genes in the mouse embryo by means of in situ hybridization. Zic genes were found as a group of genes coding for zinc finger proteins that are expressed in a restricted manner in the adult mouse cerebellum. We showed that the genes are the vertebrate homologues of Drosophila odd-paired, which may play an essential role in parasegmental subdivision and in visceral mesoderm development. The expression of the three Zic genes was first detected at gastrulation in a spatially restricted manner. At neurulation, the expression became restricted to the dorsal neural ectoderm and dorsal paraxial mesoderm. During organogenesis, the three genes were expressed in specific regions of several developing organs, including dorsal areas of the brain, spinal cord, paraxial mesenchyme, and epidermis, the marginal zone of the neural retina and distal regions of the developing limb. For all stages, significant differences in the spatial expression of Zic1, Zic2, and Zic3 were observed. Furthermore, the expression of Zic genes in Pax3, Wnt-1, and Wnt-3a mutant embryos suggested that Zic genes are not primarily regulated by the three genes which were expressed in dorsal areas similar to Zic genes. However, in open brain, a mutant with severe neural tube defects, and in the Wnt-3a mutant mice, the expression of Zic genes was changed. The changed expression pattern in Wnt-3a mutant mice suggests that Zic genes in the neural tube are regulated by the factors from notochord. Our findings suggest that Zic genes are involved in many developmental processes. Furthermore, analysis of gene expression patterns in different mouse mutants indicated that Zic genes may act upstream of many known developmental regulatory genes.     

深度学习在磁共振影像脑疾病诊断中的应用

由于脑疾病的发生会对社会产生严重危害,所以脑疾病诊断研究的重要性日益显著. 中国“脑计划”列入“十三五”规划与国务院《“健康中国2023”规划纲要》的印发表明国家对脑疾病诊疗问题的高度重视. 由于磁共振影像的高分辨率及非入侵性等优势使其成为脑疾病研究与临床检查的主要技术手段,为脑疾病诊断提供丰富的数据基础. 深度学习由于其可拓展性与灵活性在各个领域得到广泛应用,展现出巨大的发展潜力. 本文针对深度学习在典型脑疾病诊断中的应用进行综述,结构组织如下：首先对深度学习在自闭症、精神分裂症、阿尔兹海默症三种典型脑疾病诊断上的应用进行了阐述;然后对用于三种脑疾病研究的数据集和已有的开源工具进行了汇总;最后对深度学习在磁共振影像脑疾病诊断应用中的局限性及未来发展方向进行总结与展望.

     

Slow stochastic Hebbian learning of classes of stimuli in a recurrent neural network

We study unsupervised Hebbian learning in a recurrent network in which synapses have a finite number of stable states. Stimuli received by the network are drawn at random at each presentation from a set of classes. Each class is defined as a cluster in stimulus space, centred on the class prototype. The presentation protocol is chosen to mimic the protocols of visual memory experiments in which a set of stimuli is presented repeatedly in a random way. The statistics of the input stream may be stationary, or changing. Each stimulus induces, in a stochastic way, transitions between stable synaptic states. Learning dynamics is studied analytically in the slow learning limit, in which a given stimulus has to be presented many times before it is memorized, i.e. before synaptic modifications enable a pattern of activity correlated with the stimulus to become an attractor of the recurrent network. We show that in this limit the synaptic matrix becomes more correlated with the class prototypes than with any of the instances of the class. We also show that the number of classes that can be learned increases sharply when the coding level decreases, and determine the speeds of learning and forgetting of classes in the case of changes in the statistics of the input stream.     

Memory, consciousness and neuroimaging

Neuroimaging techniques that allow the assessment of memory performance in healthy human volunteers while simultaneously obtaining measurements of brain activity in vivo may offer new information on the neural correlates of particular forms of memory retrieval and their association with consciousness and intention. We consider evidence from studies with positron emission tomography and functional magnetic resonance imaging indicating that priming, a form of implicit retrieval, is associated with decreased activity in various cortical regions. We also consider evidence concerning the question of whether two components of explicit retrieval--intentional or effortful search and successful conscious recollection--are preferentially associated with increased activity in prefrontal and medial temporal regions, respectively. Last, we consider recent efforts to probe the relation between the phenomenological character of remembering and neural activity. In this instance we broaden our scope to include studies employing event-related potentials and consider evidence concerning the neural correlates of qualitatively different forms of memory, including memory that is specifically associated with a sense of self, and the recollection of particular temporal or perceptual features that might contribute to a rich and vivid experience of the past.     

Rehabilitation of brain damage: Brain plasticity and principles of guided recovery.

Rehabilitation of the damaged brain can foster reconnection of damaged neural circuits; Hebbian learning mechanisms play an important part in this. The authors propose a triage of post-lesion states, depending on the loss of connectivity in particular circuits. A small loss of connectivity will tend to lead to autonomous recovery, whereas a major loss of connectivity will lead to permanent loss of function; for such individuals, a compensatory approach to recovery is required. The third group have potentially rescuable lesioned circuits, but guided recovery depends on providing precisely targeted bottom–up and top–down inputs, maintaining adequate levels of arousal, and avoiding activation of competitor circuits that may suppress activity in target circuits. Empirical data are implemented in a neural network model, and clinical recommendations for the practice of rehabilitation following brain damage are made. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Networks of gene regulation, neural development and the evolution of general capabilities, such as human empathy

A network of gene regulation organized in a hierarchical and combinatorial manner is crucially involved in the development of the neural network, and has to be considered one of the main substrates of genetic change in its evolution. Though qualitative features may emerge by way of the accumulation of rather unspecific quantitative changes, it is reasonable to assume that at least in some cases specific combinations of regulatory parts of the genome initiated new directions of evolution, leading to novel capabilities of the brain. These notions are applied, in this paper, to the evolution of the capability of cognition-based human empathy. It is suggested that it has evolved as a secondary effect of the evolution of strategic thought. Development of strategies depends on abstract representations of one's own possible future states in one's own brain to allow assessment of their emotional desirability, but also on the representation and emotional evaluation of possible states of others, allowing anticipation of their behaviour. This is best achieved if representations of others are connected to one's own emotional centres in a manner similar to self-representations. For this reason, the evolution of the human brain is assumed to have established representations with such linkages. No group selection is involved, because the quality of strategic thought affects the fitness of the individual. A secondary effect of this linkage is that both the actual states and the future perspectives of others elicit vicarious emotions, which may contribute to the motivations of altruistic behaviour.     

Reward prediction in primate basal ganglia and frontal cortex

Reward information is processed in a limited number of brain structures, including fronto-basal ganglia systems. Dopamine neurons respond phasically to primary rewards and reward-predicting stimuli depending on reward unpredictability but without discriminating between rewards. These responses reflect 'errors' in the prediction of rewards in correspondence to learning theories and thus may constitute teaching signals for appetitive learning. Neurons in the striatum (caudate, putamen, ventral striatum) code reward predictions in a different manner. They are activated during several seconds when animals expect predicted rewards. During learning, these activations occur initially in rewarded and unrewarded trials and become subsequently restricted to rewarded trials. This occurs in parallel with the adaptation of reward expectations by the animals, as inferred from their behavioral reactions. Neurons in orbitofrontal cortex respond differentially to stimuli predicting different liquid rewards, without coding spatial or visual features. Thus, different structures process reward information processed in different ways. Whereas dopamine neurons emit a reward teaching signal without indicating the specific reward, striatal neurons adapt expectation activity to new reward situations, and orbitofrontal neurons process the specific nature of rewards. These reward signals need to cooperate in order for reward information to be used for learning and maintaining approach behavior.     

New approaches to the study of amnesic patients: what can a neurofunctional philosophy and neural network methods offer?

In this paper I first consider a neurofunctional approach to the study of amnesic patients. This approach stresses the need for theorising about the processing operations of brain regions and circuits rather than for theorising about neuropsychological syndromes. A syndrome such as amnesia-may not exist, in any meaningful sense, if there is marked heterogeneity within the patients grouped together in this way. Powerful neuroimaging techniques may now allow a more useful basis for grouping patients in terms of lesion location rather than aetiology. In turn this will allow an evaluation of the information processing functions subserved by the lesioned structures. The second strand to the present paper stresses the weakness in the specification of current theories. This has made it difficult to select experimental tasks that decisively measure the key components of those theories. The paper makes the case that explicit neural network models are a useful way to try to overcome this problem. In line with these ideas, the paper begins to build a model of how the brain may achieve useful kinds of stimulus representations. Considerations of human behaviour in category learning tasks have emphasised parallel and interacting roles for both exemplar- and element-based stimulus representations. It is suggested that the hippocampus itself may encode exemplar representations, and these may provide a basis for episodic memory as well as some types of category learning. It is further suggested that the ventral striatum may encode the element-based representations. The model allows some new and detailed predictions for the performance of amnesic subjects related to lesion location.     

Immune activation: the role of pro-inflammatory cytokines in inflammation, illness responses and pathological pain states

It has recently become accepted that the activated immune system communicates to brain via release of pro-inflammatory cytokines. This review examines the possibility that pro-inflammatory cytokines (interleukins and/or tumor necrosis factor) mediate a variety of commonly studied hyperalgesic states. We will first briefly review basic immune responses and inflammation. We will then develop the concept of illness responses and provide evidence for their existence and for the dramatic changes in neural functioning that they cause. Lastly, we will examine the potential roles that both pro-inflammatory cytokines and the neural circuits that they activate may play in the hyperalgesic states produced by irritants, inflammatory agents, and nerve damage. The possibility is raised that apparently diverse hyperalgesic states may converge in the central nervous system and activate similar or identical neural circuitry.     

On the similarity of odor and language perception

The association between olfaction and language is discussed. The effects of odor on human behavior and cognitive processing are reviewed as are electrophysiological studies of odor/language interactions. Also reviewed are specific effects of odor administration on language-dependent tasks. The hypothesis is advanced that odor information processing shares some of the cortical resources used in processing language and that interference between these two types of stimuli occurs when they are simultaneously processed. The reason for this overlap in resources is thought to be due to the similarities in the spatio-temporal patterns produced in the neural coding of odors and language.     

Neural correlates of the episodic encoding of pictures and words

A striking characteristic of human memory is that pictures are remembered better than words. We examined the neural correlates of memory for pictures and words in the context of episodic memory encoding to determine material-specific differences in brain activity patterns. To do this, we used positron emission tomography to map the brain regions active during encoding of words and pictures of objects. Encoding was carried out by using three different strategies to explore possible interactions between material specificity and types of processing. Encoding of pictures resulted in greater activity of bilateral visual and medial temporal cortices, compared with encoding words, whereas encoding of words was associated with increased activity in prefrontal and temporoparietal regions related to language function. Each encoding strategy was characterized by a distinctive activity pattern, but these patterns were largely the same for pictures and words. Thus, superior overall memory for pictures may be mediated by more effective and automatic engagement of areas important for visual memory, including medial temporal cortex, whereas the mechanisms underlying specific encoding strategies appear to operate similarly on pictures and words.     

Forecasting generalized epileptic seizures from the EEG signal by wavelet analysis and dynamic unsupervised fuzzy clustering

Dynamic state recognition and event-prediction are fundamental tasks in biomedical signal processing. We present a new, electroencephalogram (EEG)-based, brain-state identification method which could form the basis for forecasting a generalized epileptic seizure. The method relies on the existence in the EEG of a preseizure state, with extractable unique features, a priori undefined. We exposed 25 rats to hyperbaric oxygen until the appearance of a generalized EEG seizure. EEG segments from the preexposure, early exposure, and the period up to and including the seizure were processed by the fast wavelet transform. Features extracted from the wavelet coefficients were imputed to the unsupervised optimal fuzzy clustering (UOFC) algorithm. The UOFC is useful for classifying similar discontinuous temporal patterns in the semistationary EEG to a set of clusters which may represent brain-states. The unsupervised selection of the number of cluster overcomes the a priori unknown and variable number of states. The usually vague brain state transitions are naturally treated by assigning each temporal pattern to one or more fuzzy clusters. The classification succeeded in identifying several, behavior-backed, EEG states such as sleep, resting, alert and active wakefulness, as well as the seizure. In 16 instances a preseizure state, lasting between 0.7 and 4 min was defined. Considerable individual variability in the number and characteristics of the clusters may postpone the realization of an early universal epilepsy warning. University may not be crucial if using a dynamic version of the UOFC which has been taught the individual's normal vocabulary of EEG states and can be expected to detect unspecified new states.