.png){kind=icon}
.png){kind=icon}
Return
MLP iterative construction algorithm
.png){kind=icon}
PRE
AI
DOI：10.1016/S0925-2312(97)00054-4.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
This paper presents a novel multi-layer perceptron neural network architecture selection and weight training algorithm for classification problems. The MLP iterative construction algorithm (MICA) autonomously constructs an MLP neural network as it trains. Experimental results show the algorithm achieves 100% accuracy on the training data, the same or better generalization accuracies as Backprop on the test data, while using less FLOPS. Moreover, relaxation of the hidden layer nodes improves test set recognition accuracies to be greater than that of Backprop. Furthermore, seeding the Backprop algorithm with the hidden layer weights from MICA is demonstrated. The MICA seeding improves the effectiveness of Backprop and enables Backprop to solve a new class of problems, i.e., problems with areas of low mean-squared error.
Keywords:
neural network training
construction
Ho-Kashyap method
hidden node pruning
multi-layered perceptron
Journal
IF:
6.5
Papers:
2.5W
Citations:
6.5W
Organization
No organization information available