| monmlp-package {monmlp} | R Documentation |
The monmlp package implements the monotone multi-layer perceptron neural network (MONMLP) model following Zhang and Zhang (1999). The main feature is the monotone constraint, which guarantees monotonically increasing behaviour of model outputs with respect to specified covariates. The package also features model architectures with one or two hidden layers, analytical calculation of the gradient via backpropagation, and optional use of early stopping in conjunction with bootstrap aggregation to control overfitting. The model reduces to a standard multi-layer perceptron neural network if the monotone constraint is not invoked.
| Package: | monmlp |
| Type: | Package |
| License: | GPL-2 |
Alex J. Cannon <http://www.eos.ubc.ca/~acannon>
Lang, B., 2005. Monotonic multi-layer perceptron networks as universal approximators. In: W. Duch et al. (eds.): ICANN 2005, Lecture Notes in Computer Science, 3697:31-37. doi:10.1007/11550907
Minin, A., Velikova, M., Lang, B., and Daniels, H., 2010. Comparison of universal approximators incorporating partial monotonicity by structure. Neural Networks, 23:471-475. doi:10.1016/j.neunet.2009.09.002
Zhang, H. and Zhang, Z., 1999. Feedforward networks with monotone constraints. In: International Joint Conference on Neural Networks, vol. 3, p. 1820-1823. doi:10.1109/IJCNN.1999.832655
http://cran.r-project.org/web/packages/monmlp/