… We show that, if viewed through the lens of Joint SourceChannel Coding, the second term
can be reinterpreted as a power constraint term. This insight also helps set the groundwork for …

… Abstract—In this paper, we study joint source-channel coding of gaussian sources over
multiple AWGN channels where the source dimension is greater than the number of channels. …

… a novel encoder architecture for the Variational Autoencoder designed … joint source-channel
coding over AWGN channels when the source dimension m is greater than channel …

… to jointly learn the encoding and decoding processes using a new discrete variational
autoencoder … to designing the best source and channel codes, makes this joint source-channel …

… and decoder functions by two convolutional neural networks (CNNs), which are trained
jointly, and can be considered as an autoencoder with a non-trainable layer in the middle that …

… [23] proposed to combine a variational autoencoder (VAE) with a Gaussian mixture model
(GMM); however, its performance remains inferior to that of the optimization-based method. …

… sources of distortion through the optimal bit allocation, in addition to designing the best source
and channel codes, makes this joint source-channel … types of autoencoders in section 4.1. …

… Although the autoencoder-based results for both channel coding and image compression
do not yet provide significant improvements over existing standards, we highlight the fact that …

… approach is based on joint sourcechannel coding (JSCC), … , the source measurements are
directly mapped to channel sym… We apply L2 regularizer to the autoencoder model, weighted …

… Note that our usage is different than what is typical for a variational autoencoder [19], as we
are … Instead, our model is more analogous to a denoising autoencoder [20], where noise is …