This paper shows a new method for implement hybrid classifiers, they are supported by self-organizing maps and parametric approach. The first are used to data analysis by means SOM, U-Matrix and approximation to typical density functions. The space parametrization is based on marginal Box-Cox transformations. The optimal feature space is determined by maximizing the Bhattacharyya∋s distance. The effect of size samples is considered. This method has been employed in the classification of visual defects in cast aluminium. Traditional classifiers as LVQ, MLP, and systems based on rules had been implemented.
This paper shows a new method for implement hybrid classifiers, they are supported by self-organizing maps and parametric approach. The first are used to data analysis by means SOM, U-Matrix and approximation to typical density functions. The space parametrization is based on marginal Box-Cox transformations. The optimal feature space is determined by maximizing the Bhattacharyya∋s distance. The effect of size samples is considered. This method has been employed in the classification of visual defects in cast aluminium. Traditional classifiers as LVQ, MLP, and systems based on rules had been implemented. Read More
Related posts:
Providing reliability in recommender systems through Bernoulli Matrix Factorization
An alloy-agnostic machine learning framework for process mapping in laser powder bed fusion
Reduced dimensionality space of features using spectral indices for detecting changes in multitemporal Landsat-8 images
A comparison and methodological proposal for hybrid approaches to quantify environmental impacts: A case study for renewable energies
Invariant-domain preserving and locally mass conservative approximation of the Lagrangian hydrodynamics equations
CORDIC-Based Computation of Arcsine and Arccosine Functions on FPGA