Biomedical subjects

Isao Tokuda

Publications and source records attributed to Isao Tokuda.

4 recordsLinked to original sources

PubMed2003

Back-propagation learning of infinite-dimensional dynamical systems.

This paper presents numerical studies of applying back-propagation learning to a delayed recurrent neural network (DRNN). The DRNN is a continuous-time recurrent neural network having time delayed feedbacks and the back-propagation learning is to teach spatio-temporal dynamics to the DRNN. Since the time-delays make the dynamics of the DRNN infinite-dimensional, the learning algorithm and the learning capability of the DRNN are different from those of the ordinary recurrent neural network (ORNN) having no time-delays. First, two types of learning algorithms are developed for a class of DRNNs. Then, using chaotic signals generated from the Mackey-Glass equation and the Rössler equations, learning capability of the DRNN is examined. Comparing the learning algorithms, learning capability, and robustness against noise of the DRNN with those of the ORNN and time delay neural network, advantages as well as disadvantages of the DRNN are investigated.

Algorithms↗

PubMed2002

Possible neural coding with interevent intervals of synchronous firing.

Neural networks composed of excitable neurons with noise generate rich nonlinear dynamics with spatiotemporal structures of neuronal spikes. Among various spatiotemporal patterns of spikes, synchronous firing has been studied most extensively both with physiological experimentation and with theoretical analysis. In this paper, we consider nonlinear neurodynamics in terms of synchronous firing and possibility of neural coding with such synchronous firing, which may be used in the "noisy brain." In particular, reconstruction of a chaotic attractor modeling a dynamical environment is explored with interevent intervals of synchronous firing from the perspective of nonlinear time series analysis and stochastic resonance.

Journal Article↗

PubMed2002

Nonlinear analysis of irregular animal vocalizations.

Animal vocalizations range from almost periodic vocal-fold vibration to completely atonal turbulent noise. Between these two extremes, a variety of nonlinear dynamics such as limit cycles, subharmonics, biphonation, and chaotic episodes have been recently observed. These observations imply possible functional roles of nonlinear dynamics in animal acoustic communication. Nonlinear dynamics may also provide insight into the degree to which detailed features of vocalizations are under close neural control, as opposed to more directly reflecting biomechanical properties of the vibrating vocal folds themselves. So far, nonlinear dynamical structures of animal voices have been mainly studied with spectrograms. In this study, the deterministic versus stochastic (DVS) prediction technique was used to quantify the amount of nonlinearity in three animal vocalizations: macaque screams, piglet screams, and dog barks. Results showed that in vocalizations with pronounced harmonic components (adult macaque screams, certain piglet screams, and dog barks), deterministic nonlinear prediction was clearly more powerful than stochastic linear prediction. The difference, termed low-dimensional nonlinearity measure (LNM), indicates the presence of a low-dimensional attractor. In highly irregular signals such as juvenile macaque screams, piglet screams, and some dog barks, the detectable amount of nonlinearity was comparatively small. Analyzing 120 samples of dog barks, it was further shown that the harmonic-to-noise ratio (HNR) was positively correlated with LNM. It is concluded that nonlinear analysis is primarily useful in animal vocalizations with strong harmonic components (including subharmonics and biphonation) or low-dimensional chaos.

Animals↗

PubMed2001

Learning phase synchronization from nonsynchronized chaotic regimes.

We present a novel modeling approach for reconstruction of the global behavior of coupled chaotic systems from bivariate time series. We analyze two coupled chaotic oscillators, which are able to phase synchronize due to coupling. It is shown that our technique enables the recovery of the synchronization diagram from only three data sets. In particular, this allows the estimate of the relative strength of the coupling and the parameter mismatch of both subsystems. The method is most efficient if only data from the nonsynchronized regime are used for the model learning. We also apply this approach to experimental data of a paced plasma tube.

Journal Article↗