Abstract
We consider the computational power of neural networks constructed by partitioning algorithms. These neural networks can also be viewed as decision lists with tests evaluating linear functions. An exponential lower bound is proved for the complexity of an explicit Boolean function in this model. The lower bound is extended to decision trees of bounded rank. We also discuss the relationship between these models and the hierarchy of threshold circuit complexity classes.
Partially supported by NSF grant CCR-9208170.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
References
M.Anthony: Threshold functions, decision lists and the representation of Boolean functions, NeuroCOLT Tech. Rep. Ser. NC-TR-96-028, January 1996.
L.Babai, P.Frankl, J.Simon: Complexity classes in communication complexity,Proc. 27th FOCS (1986), 337 – 347.
R.P.Brent: Fast training algorithms for multilayer neural nets,IEEE Trans, on Neural Networks 2 (1991), 346 – 354.
A.Ehrenfeucht, D.Haussier: Learning decision trees from random examples,Proc. COLT’88 (1988), 182 – 194.
M.Goldmann, J.Hastad, A.A.Razborov: Majority gates vs. general weighted threshold gates,Computational Complexity 2 (1992), 277 – 300.
H.D.GrÖger, Gy.Turán: On linear decision trees computing Boolean functions,Proc. 18th ICALP, Springer LNCS 510 (1991), 707 – 718.
H.D.GrÖger, Gy.Turán: A linear lower bound for the size of threshold circuits,Bulletin of the EATCS 50 (1993), 220 – 222.
A.Hajnal, W.Maass, P.Pudlák, M.Szegedy, Gy.Turán: Threshold circuits of bounded depth,J.Comp.Syst. Sci.46 (1993), 129 – 154.
J.Hastad, M.Goldmann: On the power of small-depth threshold circuits,Computational Complexity 1 (1991), 113 – 129.
J.Hertz, A.Krogh, R.G.Palmer:Introduction to the Theory of Neural Computing. Addison-Wesley, 1991.
R.G.Jeroslow: On defining sets of vertices of the hypercube by linear inequalities,Discrete Math.11 (1975), 119 – 124.
S.A.J.Keibek, G.T.Barkema, H.M.A.Andree, M.H.F.Savenije, A.Taal: A fast partitioning algorithm and a comparison of feedforward neural networks,Europhysics Letters 18 (1992), 555 – 559.
M.Krause, S.Waack: Variation ranks of communication matrices and lower bounds for depth two circuits having symmetric gates with unbounded fan-in,Proc. 32nd FOCS (1991), 777 – 782.
M.Marchand, M.Golea: On learning simple neural concepts: from halfspace intersections to neural decision lists,Network: Computation in Neural Systems 4 (1993), 67 – 85.
M.Marchand, M.Golea, P.Ruján: A convergence theorem for sequential learning in two layer perceptrons,Europhysics Letters 11 (1990), 487 – 492.
N.J.Nilsson:The Mathematical Foundations of Learning Machines. Introduction by T.J.Sejnowski and H.White. Morgan Kaufmann, 1990.
N.Nisan: The communication complexity of threshold gates,Combinatorics, Paul Erdös is Eighty, Volume I, D.Miklós, V.T.Sós, T.Szönyi, Editors, Bolyai Math. Soc. (1993), 301 – 315.
A.A.Razborov: On small depth threshold circuits,Proc. 3rd Scandinavian Workshop on Algorithm Theory, Springer LNCS 621 (1992), 42 – 52.
R.L.Rivest: Learning decision lists,Machine Learning 2 (1987), 229 – 246.
V.Roychowdhury, K.Y.Siu, A.Orlitsky, Editors:Advances in Neural Computation and Learning. Kluwer, 1994.
P.Ruján, M.Marchand: A geometric approach to learning in neural networks,Proc. IJCNN 1989, Vol. II., 105–110. Also: Complex Systems 3 (1989), 229 – 242.
A.Schrijver:Theory of Linear and Integer Programming. Wiley, 1986.
F. Vat an: Some lower and upper bounds for algebraic decision trees and the separation problem,Proc. 7th Structure in Complexity Theory (1992), 295 – 304.
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 1997 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Turán, G., Vatan, F. (1997). Linear decision lists and partitioning algorithms for the construction of neural networks. In: Cucker, F., Shub, M. (eds) Foundations of Computational Mathematics. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-60539-0_34
Download citation
DOI: https://doi.org/10.1007/978-3-642-60539-0_34
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-61647-4
Online ISBN: 978-3-642-60539-0
eBook Packages: Springer Book Archive