Law discovery using neural networks
Published 23 August 1997
Kazumi Saito, Ryohei Nakano
Citations52
Generate an AI Snapshot to get a quick, structured summary of this paper.
Study Snapshot
ObjectiveStudy objective
MethodsResearch methodology
PopulationPopulation studied
Sample sizeSample sizes
OutcomesStudy outcomes here
ResultsStudy results comes here
LimitationsResearch study limitations comes here
A concise AI-generated summary of the paper will appear here once you click Generate AI Snapshot.
TL;DR
The main advantage of this method over previous work of symbolic or connectionist approach is that it can efficiently discover numeric laws whose power values are not restricted to integers.
Abstract
This paper proposes a new connectionist approach to numeric law discovery; i.e., neural networks (law-candidates) are trained by using a newly invented second-order learning algorithm based on a quasi-Newton method, called BPQ, and the MDL criterion selects the most suitable from law-candidates. The main advantage of our method over previous work of symbolic or connectionist approach is that it can efficiently discover numeric laws whose power values are not restricted to integers. Experiments showed that the proposed method works well in discovering such laws even from data containing irrelevant variables or a small amount of noise. 1.
Keywords
Computer SciencePhysics and Astronomy
Learning Internal Representations by Error Propagation
16,235 Citations1985David E. Rumelhart, Geoffrey E. Hinton +1 more
Mathematics and Computers in SimulationLinear and nonlinear programming.
4,850 Citations1986
The linear and nonlinear programming is universally compatible with any devices to read and is available in the book collection an online access to it is set as public so you can download it instantly.
Neural ComputationFirst- and Second-Order Methods for Learning: Between Steepest Descent and Newton's Method
1,193 Citations1992Roberto Battiti
First- and second-order optimization methods for learning in feedforward neural networks are reviewed to illustrate the main characteristics of the different methods and their mutual relations.
Medical Entomology and ZoologyScientific Discovery: Computational Explorations of the Creative Processes
812 Citations1987Pat Langley, Herbert A. Simon +2 more
Neural ComputationProduct Units: A Computationally Powerful and Biologically Plausible Extension to Backpropagation Networks
417 Citations1989Richard Durbin, David E. Rumelhart
A new form of computational unit for feedforward learning networks of the backpropagation type that calculates a weighted product, where each input is raised to a power determined by a variable weight.
International Journal of ForecastingTime series prediction: Forecasting the future and understanding the past
225 Citations1994Chris Chatfield, Andreas S. Weigend
By reading time series prediction forecasting the future and understanding the past, you can take more advantages with limited budget.
Machine LearningIntegrating Quantitative and Qualitative Discovery: The ABACUS System
182 Citations1986Brian Falkenhainer, Ryszard S. Michalski
This research attempts to integrate both types of learning by combining newly developed heuristics for formulating equations with the previously developed concept learning method embodied in the inductive learning program AQ11, resulting in ABACUS.
Neural Information Processing SystemsLearning with Product Units
58 Citations1994Laurens R. Leerink, C. Lee Giles +2 more
A constructive method is introduced which solves well-researched problems with significantly less neurons than previously reported, and some atypical heuristics to improve learning are proposed.
Neural ComputationPartial BFGS Update and Efficient Step-Length Calculation for Three-Layer Neural Networks
56 Citations1997Kazumi Saito, Ryohei Nakano
It turned out that an efficient and accurate step-length calculation plays an important role for the convergence of quasi-Newton algorithms, and a partial BFGS update greatly saves storage space without losing the convergence performance.
International Journal of Neural SystemsSUPERVISED LEARNING ON LARGE REDUNDANT TRAINING SETS
49 Citations1993Martin Möller
A new algorithm combining the good properties of off-line and on-line algorithms is introduced, which improves the efficiency of supervised learning on large redundant training sets.
Elsevier eBooksLearning Polynomial Functions by Feature Construction
27 Citations1991Richard S. Sutton, Christopher J. Matheus
A method for learning higher-order polynomial functions from examples using linear regression and feature construction and an extension to this method selected the specific pair of features to combine by measuring their joint ability to predict the hypothesis' error.
Elsevier eBooksA Robust Approach to Numeric Discovery
24 Citations1990Bernd Nordhausen, Pat Langley
The basic method of the IDS algorithm for numeric discovery is described in terms of heuristic search through a space of numeric terms, and it is shown how IDS uses the algorithm to find three classes of numeric relations.
Data-Driven Approaches to Empirical Discovery
24 Citations1988Pat Langley, Jan M. Żytkow
Machine LearningBivariate scientific function finding in a sampled, real-data testbed
18 Citations1993Cullen Schaffer
The evidence presented here supports the view that this is sometimes possible, but it also suggests how often purely data-driven discovery is not possible and how much more difficult it may be than has often been assumed.
Basis-Function Trees as a Generalization of Local Variable Selection Methods for Function Approximation
11 Citations1990Terence D. Sanger