Zobrazeno 1 - 6
of 6
pro vyhledávání: '"Shahkarami, Abtin"'
Bidirectional recurrent neural networks (bi-RNNs), in particular, bidirectional long short term memory (bi-LSTM), bidirectional gated recurrent unit, and convolutional bi-LSTM models have recently attracted attention for nonlinearity mitigation in fi
Externí odkaz:
http://arxiv.org/abs/2207.12154
Autor:
Shahkarami, Abtin
Publikováno v:
(presesntation given at 20th IEEE INTERNATIONAL CONFERENCE ON MACHINE LEARNING AND APPLICATIONS (ICMLA) (IEEE ICMLA2021), Pasadena, California, USA, 13-15 Dec 2021)
A hybrid architecture comprising a CNN encoder and a many-to-one RNN was proposed. CNN captures short-temporal dependencies, and RNN extracts long-term features. CNN lessens the computational burden on RNN by taking the data into a latent space. Than
Externí odkaz:
https://explore.openaire.eu/search/publication?articleId=doi_dedup___::47c2af2f3cba31556114d10f6f729479
https://doi.org/10.5281/zenodo.5871715
https://doi.org/10.5281/zenodo.5871715
Autor:
Shahkarami, Abtin
Publikováno v:
(presentation given at DigiCosme/GdR-ISIS: Workshop on machine learning in optical communication systems, Paris, 25/03/2021)
By leveraging a more optical parameter sharing scheme using a memory enabled by the hidden state of a recurrent layer, we propose a hybrid CNN-RNN based receiver, which achieves the BER of state-of-the-art CNN model with: 13% fewer parameters (303K v
Externí odkaz:
https://explore.openaire.eu/search/publication?articleId=doi_dedup___::0fc75f3b8e01a9108057d20f46f7ac5d
Publikováno v:
20th IEEE International Conference on Machine Learning and Applications (ICMLA 2021)
2021 20th IEEE International Conference on Machine Learning and Applications (ICMLA)
2021 20th IEEE International Conference on Machine Learning and Applications (ICMLA)
Publikováno v:
Proceedings of the Asia Communications and Photonics Conference 2021
Asia Communications and Photonics Conference 2021
Asia Communications and Photonics Conference 2021
An attention mechanism is integrated into neural network-based equalizers to prune the fully-connected output layer. For a 100 GBd 16-QAM 20 x 100 km SMF transmission, this approach reduces the computational complexity by ~15% in a CNN+LSTM model.
Publikováno v:
SN Applied Sciences; June 2019, Vol. 1 Issue: 6 p1-19, 19p