Melhoramento genético de algodoeiro colorido: Redes Neurais Artificiais versus métodos convencionais

Abstract

The objective of this work was: a) to verify the genotype x environment relationships for the physical and potential characteristics of the methods of Eberhart and Russel (1966), Centroid, as well as the use of artificial neural networks in the adaptability and stability of the cotton genotypes of b) to analyze a genetic difference between cotton genotypes of fiber and power propagation factors by the UPGMA and Tocher methods to identify the parents of potential risk factors and to evaluate the phenotypic and genotypic and indirect correlations on productivity, yield and technological characteristics of colored cotton fiber. The experiment was carried out at the experimental farm of Capim Branco, in Uberlândia-MG, in the crop year 2013/2014, 2014/2015, 2015/2016 and 2016/2017. Twelve cotton fiber genotypes were evaluated. The experimental design was completely randomized blocks with three replicates. The yield of cotton seed, fiber yield and technical characteristics of the fiber were evaluated with the aid of the HVI apparatus (High Volume Instrument), being: Average length of fiber (UHML), Uniformity of length (UI), Index of short fibers (SFI), fiber resistance (STR), fiber elongation (ELG), micronaire (MIC) and fiber maturity (MAT). GxA, which demonstrates the differential behavior of genocysts in the face of environmental oscillations. The interaction was predominantly intentional and adaptive, and a correlation was found between the Eberhart and Russell methods and the RNAs, and the genotypes UFUJP-02 and UFUJP-17 were shown to be responsive to environmental stimuli with high predictability, and to be shown to be the quality and quality of fibers. The RNA's method demonstrated how much adaptability was compared to the Eberhart and Russell and Centroid methods. Through the contribution of Singh, UHML and MAT were the characteristics that contributed most to a divergence. Five divergent groups were formed, one less than Tocher with 6 groups. Commercial applications may be more useful for the generation of segregant residues and with greater genetic variability. Aiming to increase the productive and improved potential of fiber quality, crosses between UFUJP-16 and the most frequent witnesses, the greater chance of success in the breeding program. The MIC, MAT, STR and ELG characteristics were highly comic, and the exterior was negative, that is, an inverse association with productivity. In the analysis of the MIC, MAT and STR tracks, the deleterious effects exceed the magnitude of the residual effect, and the MAT has direct effect in the unfavorable sense, demonstrating absence of cause and effect on productivity. The MIC characteristic, despite the high direct effect, has low genotype determination coefficient, making its use in indirect selection impossible. It was verified that the compromise of fiber and resistance can be used in the selection of direct, as long as a truncated selection is made between the two.