Bootstrap-inspired techniques in computational intelligence

The link between ensemble systems and bootstrap techniques can be constituted by using different training data subsets obtained by resampling of the original training data. Bootstrap resampling has been originally developed for estimating sampling distributions of statistical estimators from limited data but now finds uses in signal processing, among others. Boostrap techniques provide the information on how good an estimate is. In signal processing, it is used for signal detection and spectral estimation. In addition, boostrap-based ideas have also been used in recent development of many ensemble-based algorithms which use multiple classifiers to improve classification performance. The algorithms that are ensemble-based provides the reduction of variance and increase in confidence of a decision. Ensemble system can also be used in splitting large datasets into smaller and logical partitions. Meanwhile, the algorithm is classed into the methods of bagging and boosting. Bagging is independent of the model chosen for the individual classifer and can be used with any supervised classifier while boosting alters the training data distribution before each new bootstrap sample is obtained. Another class is adaboost which extends boosting to multiclass and regression problems. Overall, boostrap approaches make new challenges in computation intelligence to be addressable.