LSTM identification and modeling

Question

Jesus Mª Juarez Ferreras 2024-9-23

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2154890-lstm-identification-and-modeling

评论： Jesus Mª Juarez Ferreras 2024-11-5，10:48

Hello, I have a plant identified in LSTM. I have created normalized datasets, made the network architecture, (numHiddenUnits, numHiddenUnits, layers, maxEpochs, miniBatchSize, options...) Now, based on the weights of the matrices W, R and b of the trained model, I want to make the mathematical model of this plant. State space or its transfer function(s). How would this be done?

Thank you.

Hola,

tengo una planta identificada en LSTM. He creado dataset normalizados, realizado la arquitectura de la red, (numHiddenUnits, numHiddenUnits, layers,maxEpochs, miniBatchSize, options...)

Ahora, a partir de los pesos de las matrices W, R y b del modelo entrenado quiero hacer el modelo matemático de esta planta. Espacio de estados o su(s) funcione(s) de transferencia. ¿Cómo se haría esto?

Gracias.

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

praguna manvi 2024-9-30

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2154890-lstm-identification-and-modeling#answer_1524580

Hi @Jesus Mª Juarez Ferreras,

I understand that you are attempting to estimate a state-space model from a trained LSTM model. This task can be challenging for direct translation since LSTMs are complex, inherently nonlinear, and dynamic. They typically do not conform to simple empirical equations that represent the learned relationships.

However, it is possible to capture the behavior of the LSTM model and estimate an approximate “ssm” model using the “estimate” function, as detailed in this documentation:

https://www.mathworks.com/help/econ/ssm.estimate.html

Hope this helps!