CMU Artificial Intelligence Repository
Home INFO Search FAQs Repository Root

MBP: Matrix Backpropagation Package

areas/neural/systems/mbp/
MBP (Matrix Back Propagation) is an efficient implementation of the back-propagation algorithm for current-generation workstations. The algorithm includes a per-epoch adaptive technique for gradient descent. All the computations are done through matrix multiplications and make use of highly optimized C code. The goal is to reach almost peak-performances on RISCs with superscalar capabilities and fast caches. On some machines (and with large networks) a 30-40x speed-up can be measured respect to conventional implementations.
Origin:   

   risc6000.dibe.unige.it:/pub/ [130.251.89.154]
   as the files MBPv1.1.tar.Z (unix version) and
   MBPv11.zip    (DOS version)

Version: 1.1 (23-NOV-93) Requires: C, UNIX CD-ROM: Prime Time Freeware for AI, Issue 1-1 Author(s): Davide Anguita or DIBE University of Genova Via all'Opera Pia 11a 16145 Genova, ITALY Tel: +39-10-3532192 Fax: +39-10-3532175 Keywords: Authors!Anguita, Backpropagation, Gradient Descent, MBP, Machine Learning!Neural Networks, Matrix Backpropagation Package, Matrix Multiplication, Neural Networks, Univ. of Genova References: The documentation is included in the distribution as the postscript file mbpv11.ps. D.Anguita, G.Parodi, R.Zunino - An efficient implementation of BP on RISC- based workstations. Neurocomputing, in press. D.Anguita, G.Parodi, R.Zunino - Speed improvement of the BP on current generation workstations. WCNN '93, Portland. D.Anguita, G.Parodi, R.Zunino - YPROP: yet another accelerating technique for the bp. ICANN '93, Amsterdam.
Last Web update on Mon Feb 13 10:25:26 1995
AI.Repository@cs.cmu.edu