A flexible multicomputer algorithm for artificial neural networks

Abstract In this paper we test the efficiency and speedup achieved by parallel computation of a multilayer neural net algorithm, as compared to the time absorbance of a standard sequential algorithm. Analytical expressions for the communication overhead and floating point operations are derived. Based on these, the load balancing problem is solved as a nonlinear programming formulation. The parallel algorithm is tested both by time series data and cross-sectional data for financial distress prediction. The sequential version of the algorithm was coded in standard C, while the parallel version was programmed in the parallel-C language PACT by the author.

[1]  D. Rumelhart,et al.  Predicting sunspots and exchange rates with connectionist networks , 1991 .

[2]  Matthew Zeidenberg,et al.  Neural networks in artificial intelligence , 1990, Ellis Horwood series in artificial intelligence.

[3]  Richard P. Lippmann,et al.  An introduction to computing with neural nets , 1987 .

[4]  D. O. Hebb,et al.  The organization of behavior , 1988 .

[5]  Ralf Östermark.  Using Neural Nets in Modelling Vector Time Series , 1994 .

[6]  Philip D. Wasserman,et al.  Neural computing - theory and practice , 1989 .

[7]  Melody Y. Kiang,et al.  Managerial Applications of Neural Networks: The Case of Bank Failure Predictions , 1992 .

[8]  Yoh-Han Pao,et al.  Adaptive pattern recognition and neural networks , 1989 .

[9]  Chwan-Hwa John Wu,et al.  Concurrent Asynchronous Learning Algorithms for Massively Parallel Recurrent Neural Networks , 1992, J. Parallel Distributed Comput..

[10]  David F. Shanno,et al.  An interior point method for quadratic programs based on conjugate projected gradients , 1993, Comput. Optim. Appl..

[11]  Ralf Östermark,et al.  Parallel Implementation of a VARMAX Algorithm , 1994, Parallel Comput..

[12]  Masanao Aoki,et al.  State Space Modeling of Time Series , 1987 .

[13]  H. Akaike.  Markovian Representation of Stochastic Processes and Its Application to the Analysis of Autoregressive Moving Average Processes , 1974 .

[14]  Abhay B. Bulsari,et al.  System identification of a biochemical process using feed-forward neural networks , 1991, Neurocomputing.

[15]  Tarun Khanna,et al.  Foundations of neural networks , 1990 .

[16]  Ulrich Ramacher,et al.  SYNAPSE - A Neurocomputer that Synthesizes Neural Algorithms on a Parallel Systolic Engine , 1992, J. Parallel Distributed Comput..

[17]  E. Hannan,et al.  Recursive estimation of mixed autoregressive-moving average order , 1982 .