1-Norm random vector functional link networks for classification problems

Abstract This paper presents a novel random vector functional link (RVFL) formulation called the 1-norm RVFL (1N RVFL) networks, for solving binary classification problems. The solution to optimization problem of 1N is obtained by its exterior dual penalty using Newton technique. makes model robust and delivers sparse outputs, which fundamental advantage this model. output indicates that most elements in matrix are zero; hence, decision function can be achieved incorporating lesser hidden nodes compared conventional produces classifier based on smaller number input features. To put it another way, method will suppress neurons layer. Statistical analyses have been carried out several real-world benchmark datasets. proposed with two activation functions viz., ReLU sine used work. accuracies extreme learning machine (ELM), kernel ridge regression (KRR), RVFL, (K-RVFL) generalized Lagrangian twin (GLTRVFL) networks. experimental results comparable or better accuracy indicate effectiveness usability