Considering the problems that the complexity of generalized inverse limits the learning speed of extreme machine
learning(ELM), a novel ELM, called TRON-ELM, is proposed based on the trust region Newton method in which the trust
region Newton method is used to derive the output weights. The proposed method takes the Newton equation of the cost
funcion of ELM as an unconstrained optimization, and a conjugate gradient method is used to solve the equation, which
avoids solving the inverse of the Hessian matrix, thus the operation speed is improved. Meanwhile, the existence of trust
region guarantees the global convergence. The experimental results show the effectiveness of the proposed method.

extreme machine learning;trust region Newton method;conjugate gradient method;regression