Kiyohiko Nakamura

Neural mechanisms underlying the processing of Chinese words: an fMRI study.

The present study employed functional magnetic resonance imaging (fMRI) to investigate the neural mechanisms underlying orthographic, phonological and semantic processing of single character Chinese words. Twelve right-handed native Chinese speakers participated in the study. Three fundamental linguistic tasks including orthographic judgment, phonological matching and semantic association task were used. Our results demonstrated robust activation in the left posterior inferior temporal cortex (BA 37) for all three tasks. While the phonological matching task produced left-lateralized activation in the inferior frontal and parietal regions, semantic association task showed considerable bilateral activation in the inferior frontal and occipito-parietal regions. Direct comparison between phonological matching and semantic association task yielded semantic related activation in the anterior portion of the left inferior frontal gyrus (BA 47) and the right inferior frontal region (Brocas homology; BA 45). Behaviorally, there was no difference in response time between phonological matching and semantic association task. Our findings suggested that differential neural pathways were involved in the processing of meaning and sound of single-character Chinese words. The present study provided systemic information of the neural substrates underlying the processing of different components of Chinese language.

Partially observable Markov decision processes with imprecise parameters

This study extends the framework of partially observable Markov decision processes (POMDPs) to allow their parameters, i.e., the probability values in the state transition functions and the observation functions, to be imprecisely specified. It is shown that this extension can reduce the computational costs associated with the solution of these problems. First, the new framework, POMDPs with imprecise parameters (POMDPIPs), is formulated. We consider (1) the interval case, in which each parameter is imprecisely specified by an interval that indicates possible values of the parameter, and (2) the point-set case, in which each probability distribution is imprecisely specified by a set of possible distributions. Second, a new optimality criterion for POMDPIPs is introduced. As in POMDPs, the criterion is to regard a policy, i.e., an action-selection rule, as optimal if it maximizes the expected total reward. The expected total reward, however, cannot be calculated precisely in POMDPIPs, because of the parameter imprecision. Instead, we estimate the total reward by adopting arbitrary second-order beliefs, i.e., beliefs in the imprecisely specified state transition functions and observation functions. Although there are many possible choices for these second-order beliefs, we regard a policy as optimal as long as there is at least one of such choices with which the policy maximizes the total reward. Thus there can be multiple optimal policies for a POMDPIP. We regard these policies as equally optimal, and aim at obtaining one of them. By appropriately choosing which second-order beliefs to use in estimating the total reward, computational costs incurred in obtaining such an optimal policy can be reduced significantly. We provide an exact solution algorithm for POMDPIPs that does this efficiently. Third, the performance of such an optimal policy, as well as the computational complexity of the algorithm, are analyzed theoretically. Last, empirical studies show that our algorithm quickly obtains satisfactory policies to many POMDPIPs.

An intelligent data-base interface using psychological similarity between data

A question-answering system for data-base interfaces is presented which uses psychological similarity between data. The similarity relationships between data are derived from a data base that is based on a set-theoretic model of psychological similarity. These relationships are represented in the computer as a network. The generalization function enables propagation of information obtained from user similarity responses over the network. Using the generalization function, the computer determines the generic kind of question it should next pose to the user. Through this question-answering process, the knowledge-based system aids the user in specifying requests for relevant data, as well as in retrieving data from the data base. Finally, the system presented here is applied to a chemical data base, and the results of the question-answering process with the implemented system are discussed.

Firing time of neuron by interference between synaptic inputs

The relation of intensity of one input to firing time monotonically decreases: a strong input generates a spike faster while a weak one is slower. In reality, however, numerous synaptic inputs are gathered into a single neuron at various arrival times. In such a situation it is not easy to predict firing time. In this paper, we examine firing time by interference between two synaptic inputs through the Hodgkin-Huxley (HH) model, which is a well-studied standard model of spike activity. The relations of the intensity to the firing do not always monotonically decrease. We consider why this phenomenon occurs. The analysis indicates that the potassium current controlled by the variable n, which is the delayed rectifier, is crucial for the phenomenon. We also calculated the firing time in the case of the integrate-and-fire (IF) type model, in which such a rectifiable current does not flow, with the same input form as the HH model. All relations of the intensity to the firing time monotonically decrease. This is inconsistent with the case of the HH equations. This result casts doubt on patterns of firing time in the IF type model.

Neural processing in the subsecond time range in the temporal cortex

The hypothesis that cortical processing of the millisecond time range is performed by latency competition between the first spikes produced by neuronal populations is analyzed. First, theorems that describe how the mechanism of latency competition works in a model cortex are presented. The model is a sequence of cortical areas, each of which is an array of neuronal populations that laterally inhibit each other. Model neurons are integrate-and-fire neurons. Second, the model is applied to the ventral pathway of the temporal lobe, and neuronal activity of the superior temporal sulcus of the monkey is reproduced with the model pathway. It consists of seven areas: V1, V2/V3, V4, PIT, CIT, AIT, and STPa. Neural activity predicted with the model is compared with empirical data. There are four main results: (1) Neural responses of the area STPa of the model showed the same fast discrimination between stimuli that the corresponding responses of the monkey did: both were significant within 5 ms of the response onset. (2) The hypothesis requires that the response latency of cortical neurons should be shorter for stronger responses. This requirement was verified by both the model simulation and the empirical data. (3) The model reproduced fast discrimination even when spontaneous random firing of 9 Hz was introduced to all the cells. This suggests that the latency competition performed by neuronal populations is robust. (4) After the first few competitions, the mechanism of latency competition always detected the strongest of input activations with different latencies.

A theory of cerebral learning regulated by the reward system

Hypothetical mechanisms of the neocorticohippocampal system are presented. Neurophysiological and neuroanatomical findings concerning the system are integrated to demonstrate how animals associate sensory stimuli with rewarding actions: (1) cortical plasticity regulated by cholinergic/noradrenergic inputs from the hypothalamic reward system reinforces association connections between the most activated columns in the cortex; (2) the repetitive reinforcement forms association pathways connecting sensory cortical columns activated by the stimuli with motor cortical columns producing the rewarding actions; (3) after the pathways are formed, the cortex is capable of temporarily memorizing the stimuli by producing long-term potentiation through the cortico-hippocampal circuits; and (4) the memory allows the cortex to extend correct association pathways even in an environment where sensory stimuli rapidly change. A mathematical model of parts of the nervous system is presented to quantitatively examine the mechanisms. Membrane characteristics of single neurons are given by the Hodgkin-Huxley electric circuit. According to anatomical data, neural circuits of the neocortico-hippocampal system are composed by connecting populations of the model neurons. Computer simulation using physiological data concerning ion channels demonstrates how the mechanisms work and how to test the hypotheses presented.

Temporal competition as an optimal parallel processing of the cerebrohypothalamic system

Cortical processing which makes responses in a few hundred milliseconds using neurons firing at less than 50 Hz is analyzed. A mathematical model is presented to show the cortical computational architecture using a few spikes of each neuron. The cortex is represented by a sequence of areas. Each area consists of cortical columns. The columns include pyramidal cells and interneurons laterally inhibiting the cells. Membrane characteristics are given by the Hodgkin-Huxley electric circuit. Analysis of the model shows that populations of the cells encode the strength of the synaptic input into a response delay of the millisecond scale in firing ratio of the cells, even though they suffer noise and partial damage, so that the lateral inhibition in every area produces an optimal parallel processing of temporal competition where columns compete to respond faster and to escape from the inhibition. Cortical plasticity regulated by the hypothalamic reward system reinforces synaptic efficacy of connections between areas so that the competition may lead activation of sensory cortex to mononeurons producing rewarding movement.<<ETX>>

Information maximization principle explains the emergence of complex cell-like neurons

We propose models and a method to qualitatively explain the receptive field properties of complex cells in the primary visual cortex. We apply a learning method based on the information maximization principle in a feedforward network, which comprises an input layer of image patches, simple cell-like first-output-layer neurons, and second-output-layer neurons (Model 1). The information maximization results in the emergence of the complex cell-like receptive field properties in the second-output-layer neurons. After learning, second-output-layer neurons receive connection weights having the same size from two first-output-layer neurons with sign-inverted receptive fields. The second-output-layer neurons replicate the phase invariance and iso-orientation suppression. Furthermore, on the basis of these results, we examine a simplified model showing the emergence of complex cell-like receptive fields (Model 2). We show that after learning, the output neurons of this model exhibit iso-orientation suppression, cross-orientation facilitation, and end stopping, which are similar to those found in complex cells. These properties of model neurons suggest that complex cells in the primary visual cortex become selective to features composed of edges to increase the variability of the output.

Neural network model controlling saccade based on probabilistic expectation

Psychophysical experiment and analysis of the data by a linear neuronal model are presented to investigate mechanisms of expectation on probabilistic stimuli. The results suggest that the expectation is held by synaptic weight of the neuron and the neural mechanism controlling saccade based on expectation consists of two pathways working in parallel. One facilitates saccades based on expectation. The other executes saccades in response to the saccade signal. The neural mechanism is consistent with some neurophysiological findings.

Timing information in transient behavior of neuropopulations

A computational model of neuronal populations based on neurophysiological data is presented. The neuronal population consists of model neurons that simulate electrochemical processes in real neurons. Input and output of the population are defined in terms of the stochastic behavior of the single neurons. Analysis of the populational behavior shows that neuronal populations can approximately function as delay elements that transform step input to step output, and the delay intervals can specify timing information in the several-millisecond range, which is essential to high speed parallel computation. An example network composed of neuronal populations is simulated to show how the timing information can serve to achieve parallel processing in neural networks. >