Optimal feature rescaling in machine learning based on neural networks
This work addresses the specific problem of feature scaling for neural network training in industrial applications, representing an incremental improvement over existing methods.
The paper tackles the problem of improving training efficiency and generalization in feed-forward neural networks by proposing an optimal feature rescaling method using a genetic algorithm, which improved the conditioning of gradient-based training and achieved a global minimum in modeling a real industrial centerless grinding process.
This paper proposes a novel approach to improve the training efficiency and the generalization performance of Feed Forward Neural Networks (FFNNs) resorting to an optimal rescaling of input features (OFR) carried out by a Genetic Algorithm (GA). The OFR reshapes the input space improving the conditioning of the gradient-based algorithm used for the training. Moreover, the scale factors exploration entailed by GA trials and selection corresponds to different initialization of the first layer weights at each training attempt, thus realizing a multi-start global search algorithm (even though restrained to few weights only) which fosters the achievement of a global minimum. The approach has been tested on a FFNN modeling the outcome of a real industrial process (centerless grinding).