๐”– Bobbio Scriptorium
โœฆ   LIBER   โœฆ

Design and fitting of neural network transfer functions

โœ Scribed by Robert H. Schor

Publisher
Springer-Verlag
Year
1985
Tongue
English
Weight
566 KB
Volume
51
Category
Article
ISSN
0340-1200

No coin nor oath required. For personal study only.

โœฆ Synopsis

An algorithm is presented which (a) allows construction of mathematical models involving arbitrary combinations of linear cascades, parallel pathways, and feedback loops, (b) computes a total transfer function of the system, (c) performs a least-squares optimization of model parameters to best fit the model to experimental data, and (d) provides a measure of goodness-of-fit to the data. The technique has been employed to construct and test models of neural networks which mimic a class of responses observed in the cat vestibular nuclei in response to tilt, namely responses which show both a gain increase and progressive phase lag as the stimulation frequency goes from 0.01 to 2 Hz. A network consisting of a simple gain element in parallel with an inhibitory high-pass filtered version of the input provided a satisfactory fit to these data.

๐Ÿ“œ SIMILAR VOLUMES

GENERALISED TRANSFER FUNCTIONS OF NEURAL
GENERALISED TRANSFER FUNCTIONS OF NEURAL NETWORKS
โœ C.F. Fung; S.A. Billings; H. Zhang ๐Ÿ“‚ Article ๐Ÿ“… 1997 ๐Ÿ› Elsevier Science ๐ŸŒ English โš– 640 KB

When artificial neural networks are used to model non-linear dynamical systems, the system structure, which can be extremely useful for analysis and design, is buried within the network architecture. In this paper, explicit expressions for the frequency response or generalised transfer functions of

Design of function-generating mapping ne
Design of function-generating mapping networks by interactive neural-network simulation
โœ Granino A. Korn ๐Ÿ“‚ Article ๐Ÿ“… 1991 ๐Ÿ› Elsevier Science ๐ŸŒ English โš– 812 KB

We apply a new interactive simulation environment for neural-network development to the development of mapping networks, which produce learned or preset functions of real inputs. Function-mapping networks are useful for adaptive control and as general-purpose, self-learning function generators. DESI