Rational Parametrizations of Neural Networks

نویسندگان

  • Uwe Helmke
  • Robert C. Williamson
چکیده

A connection is drawn between rational functions, the realization theory of dynamical systems, and feedforward neural networks. This allows us to parametrize single hidden layer scalar neural networks with (almost) arbitrary analytic activation functions in terms of strictly proper rational functions. Hence, we can solve the uniqueness of parametrization problem for such networks.

برای دانلود رایگان متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Universal Rational Parametrizations and Toric Varieties

This note proves the existence of universal rational parametrizations. The description involves homogeneous coordinates on a toric variety coming from a lattice polytope. We first describe how smooth toric varieties lead to universal rational parametrizations of certain projective varieties. We give numerous examples and then discuss what happens in the singular case. We also describe rational ...

متن کامل

Rational Parametrizations of Real Cubic Surfaces

Real cubic algebraic surfaces may be described by either implicit or parametric equations. Each of these representations has strengths and weaknesses and have been used extensively in computer graphics. Applications involving both representations include the efficient computation of surface intersections, and triangulation of curved surfaces. One particularly useful representation is the ration...

متن کامل

ar X iv : h ep - p h / 05 05 04 4 v 2 1 2 M ay 2 00 5 The neural network approach to parton fitting

We introduce the neural network approach to global fits of parton distribution functions. First we review previous work on unbiased parametrizations of deep-inelastic structure functions with faithful estimation of their uncertainties, and then we summarize the current status of neural network parton distribution fits.

متن کامل

ar X iv : h ep - p h / 05 05 04 4 v 1 6 M ay 2 00 5 The neural network approach to parton fitting

We introduce the neural network approach to global fits of parton distribution functions. First we review previous work on unbiased parametrizations of deep-inelastic structure functions with faithful estimation of their uncertainties, and then we summarize the current status of neural network parton distribution fits.

متن کامل

Minimal generators of the defining ideal of the Rees Algebra associated to monoid parameterizations

We describe a minimal set of generators of the defining ideal of the Rees Algebra associated to a proper parametrization of any monoid hypersurface. In the case of plane curves, we recover a known description for rational parametrizations having a syzygy of minimal degree (μ = 1). We also show that our approach can be applied to parametrizations of rational surfaces having a Hilbert-Burch resol...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

عنوان ژورنال:

دوره   شماره 

صفحات  -

تاریخ انتشار 1992