Description: BP-神经网络 The neural network is trained with the Levenberg-Marquardt algorithm.
The activation functions can be either linear ( L ) or hyperbolic tangent ( H ).-Backpropagation neural network with one hidden layer for multivariate calibration. (Designed to model only one response y at a time) Platform: |
Size: 20480 |
Author:郭乐 |
Hits:
Description: The main contribution of this paper is using
optimal control theory for improving the convergence
rate of backpropagation algorithm. In the proposed
approach, the learning algorithm of backpropagation
is modeled as a minimum time control problem
in which the step-size of its learning factor is considered
as the input of this model. In contrast to the traditional
backpropagation, learning algorithms which
the step-size by trial and error, it is selected
adaptively based on optimal control criterion. The effectiveness
of the proposed algorithm is uated in
two simulations: XOR and 3-bit parity. In both simulation
examples, the proposed algorithm outperforms
well in speed and the ability to escape local minima.-The main contribution of this paper is using
optimal control theory for improving the convergence
rate of backpropagation algorithm. In the proposed
approach, the learning algorithm of backpropagation
is modeled as a minimum time control problem
in which the step-size of its learning factor is considered
as the input of this model. In contrast to the traditional
backpropagation, learning algorithms which
the step-size by trial and error, it is selected
adaptively based on optimal control criterion. The effectiveness
of the proposed algorithm is uated in
two simulations: XOR and 3-bit parity. In both simulation
examples, the proposed algorithm outperforms
well in speed and the ability to escape local minima. Platform: |
Size: 415744 |
Author:samir |
Hits:
Description: Biological systems are able to recognise temporal sequences of stimuli or compute
in the temporal domain. In this paper we are exploring whether a biophysical
model of a pyramidal neuron can detect and learn systematic time
delays between the spikes from dierent input neurons. In particular, we investigate
whether it is possible to reinforce pairs of synapses separated by a
dendritic propagation time delay corresponding to the arrival time dierence
of two spikes from two dierent input neurons. We examine two subthreshold
learning approaches where the rst relies on the backpropagation of EPSPs (excitatory
postsynaptic potentials) and the second on the backpropagation of a
somatic action potential, whose production is supported by a learning-enabling
background current. Platform: |
Size: 458752 |
Author:nabill |
Hits: