This paper presents a comparative study of two sampling techniques for training neural networks committees. The sampling techniques used are: bootstrapping and half&half bagging. The following aggregation schemes used are employed in the study: 1) weighted averaging, 2) combination via Choquet integral with the lamda-fuzzy measures, 3) linear combination of order statistics, 4) combination via Choquet integral with the 2-additive fuzzy measure and 5) combination via Choquet integral with ordinary fuzzy measure.