A complex-valued RTRL algorithm for recurrent neural networks. (English) Zbl 1062.68098
Summary: A Complex-valued Real-Time Rercurrent Learning (CRTRL) algorithm for the class of nonlinear adaptive filters realized as fully connected recurrent neural networks is introduced. The proposed CRTRL is derived for a general complex activation function of a neuron, which makes it suitable for nonlinear adaptive filtering of complex-valued nonlinear and nonstationary signals and complex signals with strong component correlations. In addition, this algorithm is generic and represents a natural extension of the real-valued RTRL. Simulations on benchmark and real-world complex-valued signals support the approach.
References:
| [1] | DOI: 10.1109/72.279181 · doi:10.1109/72.279181 |
| [2] | DOI: 10.1109/78.127967 · doi:10.1109/78.127967 |
| [3] | Coelho P. H. G., International Joint Conference on Neural Networks 1 pp 120– (2001) |
| [4] | DOI: 10.1016/0364-0213(90)90002-E · doi:10.1016/0364-0213(90)90002-E |
| [5] | Feldkamp L. A., IEEE Transactions on Neural Networks 86 (4) pp 2259– (1998) |
| [6] | DOI: 10.1109/82.142037 · Zbl 0825.94309 · doi:10.1109/82.142037 |
| [7] | DOI: 10.1109/78.348134 · doi:10.1109/78.348134 |
| [8] | DOI: 10.1049/el:19921186 · doi:10.1049/el:19921186 |
| [9] | DOI: 10.1109/82.279210 · doi:10.1109/82.279210 |
| [10] | DOI: 10.1162/089976603321891846 · Zbl 1085.68631 · doi:10.1162/089976603321891846 |
| [11] | DOI: 10.1109/78.134446 · doi:10.1109/78.134446 |
| [12] | DOI: 10.1109/72.80202 · doi:10.1109/72.80202 |
| [13] | DOI: 10.1162/neco.1993.5.2.165 · doi:10.1162/neco.1993.5.2.165 |
| [14] | DOI: 10.1162/neco.1993.5.2.165 · doi:10.1162/neco.1993.5.2.165 |
| [15] | DOI: 10.1109/72.279191 · doi:10.1109/72.279191 |
| [16] | DOI: 10.1109/72.279187 · doi:10.1109/72.279187 |
| [17] | DOI: 10.1109/PROC.1975.9802 · doi:10.1109/PROC.1975.9802 |
| [18] | DOI: 10.1162/neco.1989.1.2.270 · doi:10.1162/neco.1989.1.2.270 |
This reference list is based on information provided by the publisher or from digital mathematics libraries. Its items are heuristically matched to zbMATH identifiers and may contain data conversion errors. In some cases that data have been complemented/enhanced by data from zbMATH Open. This attempts to reflect the references listed in the original paper as accurately as possible without claiming completeness or a perfect matching.
