An Effective Online Sequential Stochastic Configuration Algorithm for Neural Networks

Random Vector Functional-link (RVFL) networks, as a class of random learner models, have received careful attention from the neural network research community due to their advantages in obtaining fast learning algorithms and which hidden layer parameters are randomly generated remain fixed during training phase. However, its universal approximation ability may not be guaranteed if properly selected an appropriate range. Moreover, resulting learner’s generalization performance seriously deteriorate once RVFL network’s structure is well-designed. Stochastic configuration (SC) algorithm, incrementally constructs approximator by under specified supervisory mechanism, instead fixing selection scope advance without any reference information, can effectively circumvent these awkward issues caused randomness. This paper extends SC algorithm online sequential version, termed OSSC means recursive least square (RLS) technique, aiming copy with modeling tasks where observations sequentially provided. Compared networks (OS-RVFL short), our proposed avoid setting certain unreasonable range for parameters, also successfully build preferable capabilities. The experimental study has shown effectiveness algorithm.