Neural network tomography: Network replication from output surface geometry

Authors:
Rupert C. J. Minnett;Andrew T. Smith;William C. Lennon, Jr.;Robert Hecht-Nielsen
Affiliations:
Electrical and Computer Engineering Department, University of California, San Diego, La Jolla, CA 92093-0407, United States;Computer Science and Engineering Department, University of California, San Diego, La Jolla, CA 92093-0404, United States;Electrical and Computer Engineering Department, University of California, San Diego, La Jolla, CA 92093-0407, United States;Electrical and Computer Engineering Department, University of California, San Diego, La Jolla, CA 92093-0407, United States
Venue:
Neural Networks
Year:
2011

Citing 11
Cited 2

Multilayer feedforward networks are universal approximators

Neural Networks
Original Contribution: Uniqueness of the weights for minimal feedforward nets with a given input-output map

Neural Networks
On the geometry of feedforward neural network error surfaces

Neural Computation
Natural gradient works efficiently in learning

Neural Computation
Local minima and plateaus in hierarchical structures of multilayer perceptions

Neural Networks
Curves and surfaces for CAGD: a practical guide

Curves and surfaces for CAGD: a practical guide
Talking Nets: An Oral History of Neurocomputing

Talking Nets: An Oral History of Neurocomputing
The design of self-organizing polynomial neural networks

Information Sciences—Informatics and Computer Science: An International Journal
Functionally equivalent feedforward neural networks

Neural Computation
The multilayer perceptron as an approximation to a Bayes optimal discriminant function

IEEE Transactions on Neural Networks
Neural network classification: a Bayesian interpretation

IEEE Transactions on Neural Networks

Evolutionary algorithm assisted by surrogate model in the framework of ordinal optimization and optimal computing budget allocation

Information Sciences: an International Journal
Fast and Stable Learning Utilizing Singular Regions of Multilayer Perceptron

Neural Processing Letters

Quantified Score

Hi-index	0.00

Visualization

Abstract

Multilayer perceptron networks whose outputs consist of affine combinations of hidden units using the tanh activation function are universal function approximators and are used for regression, typically by reducing the MSE with backpropagation. We present a neural network weight learning algorithm that directly positions the hidden units within input space by numerically analyzing the curvature of the output surface. Our results show that under some sampling requirements, this method can reliably recover the parameters of a neural network used to generate a data set.