Hybrid sequential Monte Carlo/Kalman methods to train neural networks in non-stationary environments.

Joo F. G. de Freitas, Mahesan Niranjan, Andrew H. Gee

Browse the full ICASSP paper archive.