Error Using Prediction With LSTM
2 次查看(过去 30 天)
显示 更早的评论
Uriel Eduardo Carrero Cuadrado
2021-6-23
评论: Uriel Eduardo Carrero Cuadrado
2021-7-10
Hi Everyone!
I trained an LSTM with Keras and I'm importing this network with a .h5 file and it has the next characteristics:
![](https://www.mathworks.com/matlabcentral/answers/uploaded_files/662685/image.png)
Dimensions for inputs in this network with keras are a 3D matrix composed by (number of samples, time steps, number features per time step), I'm trying the same dimension in MATLAB but I get this error:
Undefined function 'predict' for input arguments of type 'nnet.cnn.layer.Layer'.
This is the code that I'm using for predict:
modelfile = 'Modelos\ControlLSTM_XYZ_1.h5';
net = importKerasLayers(modelfile)
save('Modelos\netLSTM.mat','net')
Example=randi([0 10],5,4,24)/10;
predict(net,Example)
In this case 'Example' is a matrix of inputs with random values between 1 and 0, that I'd use for generate a prediction with my neural network, I'd like to know if this format for inputs change in MATLAB or if it still the same, due to it's a network from Keras, also I'd like to know if in my code I'm doing some mistake with LSTM networks because I don't understanf why MATLAB show an error related with convolutional neural networks?.
Thanks for your help.
0 个评论
采纳的回答
Swetha Polemoni
2021-7-5
编辑:Swetha Polemoni
2021-7-5
Hi
It is observed that you are calling "predict" on the layerGraph object/layers array. predict is not allowed on layerGraph object/layers array. Before calling predict with layerGraph object, the layerGraph object has to be converted to dagnetwork using assembleNetwork. You can find an eample of this case in the following documentation Assemble Network from Pretrained Keras Layers.
To understand the the difference between LayerGraph and DAGNetwork refer to the following answerwhat is the difference between LayerGraph and DAGNetwork in deep learning?
更多回答(0 个)
另请参阅
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!