A comparison of three pruning algorithms for feedforward neural networks design is proposed, namely MOBD (Modified Optimal Brain Damage), TOBD (Tridiagonal Optimal Brain Damage) and OBD (Optimal Brain Damage). The application taken into account is nonlinear regression and the evaluation criteria are: mean square error, peak signal-to-noise ratio, total number of parameters in the final neural structure, weight values, number of "multiply + add" operations in the test phase and required test time.
Feedforward neural network, pruning method, saliency, symmetry.
All articles published by Studies in Informatics and Control (SIC) are published in accordance with the terms and conditions of the Creative Commons Attribution License (CC BY-NC-SA 4.0).