I want to design a LSTM network with 25 input units(T_x=25), and 5 out units(T_y=5), but I don't know how to convert 25 y^s(y_hat), to 5 y^s?

For example the following picture shows a LSTM network with equal input and outputs(equal to Tx). But what if I have 25 inputs and need 5 outputs not 25! enter image description here

Should I add the output of the first unit(y<1>) to the input of the second unit(x<2>)? Or maybe adding it to the activation or memory of the second unit(a or c)? Or I must forget the 20 first outputs(y<1> to y<20>) and only use the last 5 outputs(y<21> to y<25>)?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.