Gradient boosting works by building a sequence of models that are fitted iteratively to the data. For any step, a model is fitted to the data and residuals are obtained. A new model is fitted to those residuals, and the new model is joined to the previous one by summing them using an appropriate weight. 

The SMOTE algorithm, presented here, works differently by finding the K-nearest neighbors for each sample in the small group. Then one of those is picked up as a reference point, and random numbers between 0 and 1 are drawn and multiplied by that vector of features that was picked.