Retraining Deep denoising Autoencoder

Hello,

How Can I retrain composed two DDAEs. where first and second DDAEs have different window lengths of one and three frames respectively. Noisy speech features are used as the input of the first DDAE and its output, along with one past and one future enhanced frames from outputs of the first DDAE, are given to the next DDAE whose window length would be three.

Thanks a lot

Answer 1

[BERGHOUT Tarek]

you will nerver long use the speach frames in the second DAEs,

-first: you will encode your input frames with the first DAs.

-second:you will use the encoded inputs (the new representation of the inputs(hodden layers)) as the input of the next DEs.

you can mange my code to be capable to train stacked DAEs , it is very fast, if you liked it please rate it and give us your opinion .

https://www.mathworks.com/matlabcentral/fileexchange/71115-denoising-autoencoders?s_tid=prof_contriblnk