A linear-programming method inspired by the neural networks framework
J.C. Culioli, V. Protopopescu · OSTI OAI (U.S. Department of Energy Office of Scientific and Technical Information) · 1988
We propose a discrete numerical algorithm for linear programming that has common features with nonlinear optimization algorithms and with neural network models. The ''computation energy'' of our algorithm is not quadratic, but linear, which greatly simplifies the model. The collective computational properties are obtained by dividing the neurons in two classes: primal and dual. The algorithm is parallelisable, easier to code than most classical linear-programming algorithms, and implementable in YLSI. The relationship with other neural network models is discussed.