Showing papers in "Neurocomputing in 2001"

TL;DR: A new technique for blind source separation of speech signals by applying the decorrelation method proposed by Molgedey and Schuster in the time–frequency domain based on the temporal structure ofspeech signals is introduced.

TL;DR: A number of significant hybrid models for ASR are reviewed, putting together approaches and techniques from a highly specialistic and non-homogeneous literature, allowing for tangible improvements in recognition performance over the standard HMMs in difficult and significant benchmark tasks.

TL;DR: Rank Order Coding is an alternative to conventional rate coding schemes that uses the order in which a neuron's inputs fire to encode information, and orientation like receptive fields arise in neurons despite the fact that the input neurons never fired more than once.

TL;DR: It is shown that associative memory networks with small-world architectures can provide the same retrieval performance as randomly connected networks while using a fraction of the total connection length.

TL;DR: A number of fundamental issues are tackled including specificity of information granules vis-a-vis learning complexity in the neural networks along with their generalization features.

TL;DR: The numerical experiments show the ability of rough sets to select reduced set of pattern's features (minimizing the pattern size), while providing better generalization of neural-network texture classifiers.

TL;DR: This paper proposes an alternative approach to order the components according to their joint contributions in data reconstruction, which naturally leads the component ordering to a typical combinatorial optimization problem, whereby the underlying optimum ordering can be found in an exhaustive way.

TL;DR: The basic principles of regularisation theory are outlined for both linear and nonlinear training and then extended to cover a new hybrid training algorithm for feedforward neural networks recently proposed by the authors.

TL;DR: Fuzzy neural networks are connectionist systems that facilitate learning from data, reasoning over fuzzy rules, rule insertion, rule extraction, and rule adaptation, and EFuNNs are suitable for fast learning of on-line incoming data, adaptive learning of speech and video data, incremental learning and knowledge discovery from large databases.

TL;DR: A theory to calculate the effect of magnetic stimulation on arbitrary neuronal structure is developed and a computer simulation shows that a single magnetic pulse applied to model cortical neurons can induce brief burst firing followed by silent period of duration comparable to experimental data of TMS.

TL;DR: This work proposes a model avoiding iterative computation by using the fact that real neurons in the retina or LGN fire asynchronously, with the most activated firing first, and shows that contour integration can be obtained in a purely feed-forward way.

TL;DR: It is claimed that the same techniques can be applied to the pruning problem, and thus they are a useful tool for compaction of information.

TL;DR: In order to overcome the discontinuity of the inverse kinematics function, a novel modular neural network architecture is proposed in this paper and it is proposed that this architecture can be implemented as a continuous neural network.

TL;DR: A model of elongation is presented in which tubulin is produced in the soma and transported to the growth cone by diffusion and active transport and shows that cessation of growth in one neurite—e.g., when it encounters a target—can trigger the growth of the other neurites.

TL;DR: It is suggested that calcium-mediated rebound mechanisms intrinsic to cerebral cortex can explain the genesis of intracortical SW activity.

TL;DR: This work uses lead field theory to quantify the spatial resolution of scalp EEG, contrasting two electrode spacings and two values of skull conductivity, and shows that, without cortical constraints, 19-electrode EEG systems have optimal spatial resolution near 22– 37 cm 3, while 129-Electrode systems have 6– 8 cm 3.

TL;DR: Simulation results demonstrate that the competitive Hop"eld neural network imposed on the proposed energy function ensures an appropriate approach to solving this class of scheduling problems.

TL;DR: This research considers that the error saturation (ES) condition, which is caused by the use of gradient descent method, will greatly slow down the learning speed of BP algorithm, and proposes an error saturation prevention (ESP) function to prevent the nodes in the output layer from the ES condition.

TL;DR: The architectures of fuzzy-rough and rough-fuzzy subnets, a more comprehensive introduction to serial combinations of rough and neofuzzy neurons in neurocomputing, are described.

TL;DR: A navigation system for robotic manipulators that combines a repelling influence related to the distance between manipulator and nearby obstacles with the attracting influence produced by the angular difference between actual and final manipulator configuration to generate actuating motor commands.

TL;DR: A new thresholding method for the multiresolution decomposition of the multichannel data based on the discrete wavelet transform that can reduce both spatially correlated and uncorrelated noise components in the neural recordings.

TL;DR: It is shown that adaptation based on sensory feedback permits a stable gait even in an underactuated condition: the leg can be driven using a hip actuator alone while the knee is purely passive.

TL;DR: The results show that the partial somatic reset mechanism is the most likely candidate to reflect the mechanism used in the brain for reproducing irregular firing.

TL;DR: An efficient simplified model is proposed which preserves voltage attenuation for somatic current injection as well as for distributed synaptic inputs in dendrites and was correctly captured in the presence of synaptic background activity.

TL;DR: This article presents a generative model for spike time dependent plasticity based on a simplified model of the synaptic kinetic, and extends this model to a simplified models of integrate-and-fire neurons network using rank order coding.

TL;DR: A synaptic plasticity model where potentiation and depression are induced by precisely timed pairs of synaptic events and postsynaptic spikes is explored, including the observation that strong synapses undergo relatively less potentiation than weak synapses.

TL;DR: The properties of cells in the prefrontal cortex and inferotemporal cortex recorded in monkeys performing delayed matching-to-sample tasks with intervening visual stimuli and memory guided attention tasks are reproduced by means of a model in which two networks of leaky integrate-and-fire neurons representing the two cortical areas interact reciprocally.

TL;DR: It is shown that despite the background, volleys of precisely synchronized action potentials can stably propagate within a model network of basic integrate-and-fire neurons.

TL;DR: Axonal pathfinding is an important feature of embryonary development and both motor and sensorial function of neurite outgrowth were modelled, and assumptions were made on how filopodia could sense a gradient and be reorganized in an asymmetric distribution.

TL;DR: A model containing transient Na+ current (INa), leakage and outward currents can generate oscillations with long plateau duration and the plateaus arise from a window mode of INa and do not require noninactivating Na+Current (IP).