QFWEC ensemble method

An ensemble method that employs quadradic form to seek an optimal weight vector of classifiers

'weighted classifier ensemble based on quadratic form' : this method weights each classifier in an ensemble by maximising three different quadratic forms. This method seeks the optimal weight vector of classifiers directly by minimising the ensemble error and avoid efficiently the difficulty derived of balancing diversity and accuracy. Based on two constraints, we approximate the normal ensemble error in an an objective function. By introducing an initial weight vector of classifiers, the minimization of the objective function is equivalent to the maximization of quadratic forms. In fact, when the value of the quadratic form is larger, the obtained ensemble error corresponding to it is smaller, and also smaller than the ensemble error corresponding to the initial weight vector. Particularly, in QFWEC method, the initial weight vector can be given by user, and it can also given by a kind of ensemble methods. It means this method can improve the performance of other ensemble methods, when the weight vector obtained by other ensemble methods is used as the initial weight vector of the QFWEC method. http://www.sciencedirect.com/science/article/pii/S0031320314004348.