K. Doya and A.I. Selverston. Dimension reduction of biological neuron models by artificial neural networks. Neural Computation, 6:696--717, 1994.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....It also accounts for the decreasing density shape of firing rates that has been reported in the literature. 1 Introduction The reduction of detailed conduction based models of neuronal firing into simpler, lowerdimensional descriptions has recently received considerable attention (e.g. Doya and Selverston, 1994, Ermentrout, 1994 ] Such reductions lead to frequency current curves similar to the curve obtained with the more detailed Hodgkin Huxley model [ Doya and Selverston, 1994 ] The shape of resulting frequency current curves, which can be viewed as input output activation functions governing ....

.... models of neuronal firing into simpler, lowerdimensional descriptions has recently received considerable attention (e.g. Doya and Selverston, 1994, Ermentrout, 1994 ] Such reductions lead to frequency current curves similar to the curve obtained with the more detailed Hodgkin Huxley model [ Doya and Selverston, 1994 ] The shape of resulting frequency current curves, which can be viewed as input output activation functions governing the neuronal firing rate, can qualitatively be described as Threshold Sigmoid (TS) That is, a neuron fires only when its integrated input current (denoted also as input field) ....

K. Doya and A.I. Selverston. Dimension reduction of biological neuron models by artificial neural networks. Neural Computation, 6:696--717, 1994.

....and Huxley leads to a model based estimator e.g. of the membran potential. Other model free estimators of neuronal activity are e.g. the method of approximating the output spike train by applying a linear filter procedure on the stimulus train (Gabbiani, 1996) or the neural network approach of (Doya and Selverston, 1994) fitting a complex biological neuron by a recurrent multilayer network. The common characteristics of the mentioned model free estimators is to see the neuron in a first instance as a black box and to catch the main characteristics of the input output relationship in terms of there specific ....

Doya, K. & Selverston, A. I. (1994). Dimension Reduction of Biological Neuron Models by Artificial Neural Networks. Neural Computation, 6(4):696--717.

....during flight. It is argued that the weight matrices of such neural networks may make the analysis of biological data more approachable. In other work, artificial neural networks have served to reduce the dimensionality of high dimensional dynamical systems, such as conductance based neuron models [Doya and Selverston, 1994]. After dimensionality reduction the models are more amenable to mathematical analysis and numerical simulations. In the remainder of this paper, I will discuss the results of a pilot study in using artificial neural networks as an alternative to compartmental models. In this particular ....

Doya, Kenji and Selverston, Allen, I. (1994): "Dimension Reduction of Biological Neuron Models by Artificial Neural Networks," Neural Computation, vol. 6, no. 4, 696-717.

....five units. This five dimensional data was used to train a feed forward network to recognize the identity of the subjects. 120 images were used in training and 40 for testing. The paper reports a classification accuracy of 98 on the training set and 95 on the test set. A study was conducted in [4] on dimensionality reduction of biological neuron models which belong to the class of dissipative dynamical systems, using bottleneck neural networks. The authors state that trajectories of dissipative dynamical systems usually fall into some low dimensional manifold of the state space after some ....

....the studies done in the area of NLPCA using 1. bottleneck architectures and complex sigmoidal nodes to store phase information, 2. nonlinear extensions to unsupervised Hebbian learning, 3. INCA. These algorithms, though more complex than the PCA, are a definite improvement over it as shown in [3] [4], 8] and [14] The next chapter focuses on the bottleneck architectures and their effectiveness in 37 reducing the dimensionality of well defined signals. Chapter 3 presents the experiments conducted using bottleneck architectures for feature extraction from complex EEG signals. Chapter 3 ....

Kenji Doya. Dimension reduction of biological neuron models by artificial neural networks. Neural Computation, 6:696--717, 1994.

....It also accounts for the decreasing density shape of firing rates that has been reported in the literature. 1 Introduction The reduction of detailed conduction based models of neuronal firing into simpler, lowerdimensional descriptions has recently received considerable attention (e.g. Doya and Selverston, 1994, Ermentrout, 1994 ] Such reductions lead to frequency current curves similar to the curve obtained with the more detailed Hodgkin Huxley model [ Doya and Selverston, 1994 ] The shape of resulting frequency current curves, which can be viewed as input output activation functions governing ....

.... models of neuronal firing into simpler, lowerdimensional descriptions has recently received considerable attention (e.g. Doya and Selverston, 1994, Ermentrout, 1994 ] Such reductions lead to frequency current curves similar to the curve obtained with the more detailed Hodgkin Huxley model [ Doya and Selverston, 1994 ] The shape of resulting frequency current curves, which can be viewed as input output activation functions governing the neuronal firing rate, can qualitatively be described as Threshold Sigmoid (TS) That is, a neuron fires only when its integrated input current (denoted also as input field) ....

K. Doya and A.I. Selverston. Dimension reduction of biological neuron models by artificial neural networks. Neural Computation, 6:696--717, 1994.

....analysis of the system. 5 Discussion The present fuzzy method considers the cell membrane as a black box and directly rebuilds its input output behavior. In this sense the fuzzy system is a model free estimator. Another modelfree estimator is obtained by the artificial neural network approach of (Doya and Selverston, 1994) which fits the membrane potential of a complex biological neuron by a recurrent multilayer network. Our fuzzy logic approach combines the power of fitting complex data from artificial neural networks and the possibility of structuring the knowledge by linguistic rules from fuzzy logic. A ....

Doya, K. & Selverston, A. I. (1994). Dimension Reduction of Biological Neuron Models by Artificial Neural Networks. Neural Computation, 6(4):696--717.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC