RNN BCJR: a fully trainable version of the additive BCJR algorithm

التفاصيل البيبلوغرافية
العنوان: RNN BCJR: a fully trainable version of the additive BCJR algorithm
المؤلفون: Montorsi, Guido, Ripani, Barbara
المساهمون: Montorsi, Guido, Ripani, Barbara
بيانات النشر: IEEE
سنة النشر: 2023
المجموعة: PORTO@iris (Publications Open Repository TOrino - Politecnico di Torino)
مصطلحات موضوعية: Additive SISO, BCJR algorithm, model-driven recurrent neural networks
الوصف: We present a new version of the additive BCJR algorithm based on a recurrent neural network whose structure reflects an underlying trellis diagram. Starting from a matrix version of the equations of the additive BCJR algorithm, we derive the equivalent trainable recurrent neural network model, named Recurrent Neural Network (RNN) BCJR. The RNN BCJR consists of a linear layer to form the edge metrics from the state and input metrics, followed by a SOFTMAX/max* layer to marginalize the edge metrics back to the state and output spaces. We derive the recursions for delta propagation to train the two-layer mixing matrices from the output cost function. Unlike the previous approaches, the proposed RNN BCJR can completely replace the BCJR and is trainable from the cost functions of the outputs. The trained RNN BCJR achieves the same optimal performance as the BCJR when the model is known but at the same time can adapt itself to model mismatch, thus outperforming BCJR.
نوع الوثيقة: conference object
وصف الملف: ELETTRONICO
اللغة: English
Relation: info:eu-repo/semantics/altIdentifier/isbn/978-1-5386-7462-8; info:eu-repo/semantics/altIdentifier/wos/WOS:001094862603133; ispartofbook:International Conference on Communications; International Conference on Communications (ICC); numberofpages:6; https://hdl.handle.net/11583/2981777; info:eu-repo/semantics/altIdentifier/scopus/2-s2.0-85178253440; https://ieeexplore.ieee.org/document/10279315
DOI: 10.1109/ICC45041.2023.10279315
الاتاحة: https://hdl.handle.net/11583/2981777
https://doi.org/10.1109/ICC45041.2023.10279315
https://ieeexplore.ieee.org/document/10279315
Rights: info:eu-repo/semantics/closedAccess
رقم الانضمام: edsbas.5019E3C3
قاعدة البيانات: BASE
الوصف
DOI:10.1109/ICC45041.2023.10279315