Note to users. If you're seeing this message, it means that your browser cannot find this page's style/presentation instructions -- or possibly that you are using a browser that does not support current Web standards. Find out more about why this message is appearing, and what you can do to make your experience of our site the best it can be.

Site Tools

  • AAAS
  • Subscribe
  • Feedback

Site Search

Search Advanced

Science 23 February 1990:
Vol. 247. no. 4945, pp. 978 - 982
DOI: 10.1126/science.247.4945.978
Prev | Table of Contents | Next

Articles

Regularization Algorithms for Learning That Are Equivalent to Multilayer Networks

T. Poggio 1 and F. Girosi 1

1 Artificial Intelligence Laboratory, Center for Biological Information Processing, Massachusetts Institute of Technology, Cambridge, MA 02139

Learning an input-output mapping from a set of examples, of the type that many neural networks have been constructed to perform, can be regarded as synthesizing an approximation of a multidimensional function (that is, solving the problem of hypersurface reconstruction). From this point of view, this form of learning is closely related to classical approximation techniques, such as generalized splines and regularization theory. A theory is reported that shows the equivalence between regularization and a class of three-layer networks called regularization networks or hyper basis functions. These networks are not only equivalent to generalized splines but are also closely related to the classical radial basis functions used for interpolation tasks and to several pattern recognition and neural network algorithms. They also have an interesting interpretation in terms of prototypes that are synthesized and optimally combined during the learning stage.

Submitted on July 27, 1989
Accepted on November 29, 1989


THIS ARTICLE HAS BEEN CITED BY OTHER ARTICLES:
MSTd Neuronal Basis Functions for the Population Encoding of Heading Direction.
S. Ben Hamed, W. Page, C. Duffy, and A. Pouget (2003)
J Neurophysiol 90, 549-558
   Abstract »    Full Text »    PDF »
Auditory Space-Time Receptive Field Dynamics Revealed by Spherical White-Noise Analysis.
R. L. Jenison, J. W. H. Schnupp, R. A. Reale, and J. F. Brugge (2001)
J. Neurosci. 21, 4408-4415
   Abstract »    Full Text »    PDF »
Learning of Visuomotor Transformations for Vectorial Planning of Reaching Trajectories.
J. W. Krakauer, Z. M. Pine, M.-F. Ghilardi, and C. Ghez (2000)
J. Neurosci. 20, 8916-8924
   Abstract »    Full Text »    PDF »
Spatial Generalization from Learning Dynamics of Reaching Movements.
R. Shadmehr and Z. M. K. Moussavi (2000)
J. Neurosci. 20, 7807-7815
   Abstract »    Full Text »    PDF »
Quantization of continuous arm movements in humans with brain injury.
H. I. Krebs, M. L. Aisen, B. T. Volpe, and N. Hogan (1999)
PNAS 96, 4645-4649
   Abstract »    Full Text »    PDF »
Interpreting Neuronal Population Activity by Reconstruction: Unified Framework With Application to Hippocampal Place Cells.
K. Zhang, I. Ginzburg, B. L. McNaughton, and T. J. Sejnowski (1998)
J Neurophysiol 79, 1017-1044
   Abstract »    Full Text »    PDF »
Neural Networks in Clinical Medicine.
W. Penny and D. Frost (1996)
Med Decis Making 16, 386-398
   Abstract »    PDF »
A Theory of How the Brain Might Work.
T. Poggio (1990)
Cold Spring Harb Symp Quant Biol 55, 899-910
   Abstract »    PDF »


To Advertise     Find Products

Science. ISSN 0036-8075 (print), 1095-9203 (online)