Implementation of an an Autoencoder of wireless physical layer using deep learning in Keras/Tensorflow/pytorch
Budget: $10 – $30 USD
You need to implement the following case:
Input to AE 1: x1,
Input to AE 2: x2
signal 1= x1+noise+x2
signal 2 = x2+noise+x1
Autoencoder 1 should decode x1 successfully.
For that you need to extract crossentropy loss , and build a weight fucntion alpha + loss1/loss1+loss 2.
Train both AEs jointly and assign the weigh to loss of both such that L1*alpha+L2*1-alpha.
so that AE when alpha is 1 AE 1 learns to decode x1 and when alpha is 0 Ae 2 learns to decode x2. Rest we will discuss if you are interested in the project,
Input to AE 1: x1,
Input to AE 2: x2
signal 1= x1+noise+x2
signal 2 = x2+noise+x1
Autoencoder 1 should decode x1 successfully.
For that you need to extract crossentropy loss , and build a weight fucntion alpha + loss1/loss1+loss 2.
Train both AEs jointly and assign the weigh to loss of both such that L1*alpha+L2*1-alpha.
so that AE when alpha is 1 AE 1 learns to decode x1 and when alpha is 0 Ae 2 learns to decode x2. Rest we will discuss if you are interested in the project,