IEEE Transactions on Information Theory
System identification
Hi-index | 0.00 |
Set membership (SM)-based techniques, with least square error overlay, suffer from a trade-off between interpretability and proof of convergence. The authors introduce a modified SM algorithm with 'forgetting' covariance updating in conjunction with minimum volume data selecting strategy. The convergence properties of this algorithm and its resemblance to the stochastic approximation method are discussed.