The paper introduces several novel applications of deep learning (DL) in the physical layer of communication systems. By treating a communications system as an autoencoder, the authors propose a new approach to design end-to-end reconstruction tasks that jointly optimize transmitter and receiver components. This concept is extended to multiple transmitters and receivers, leading to the introduction of radio transformer networks (RTNs) which integrate domain-specific knowledge into the machine learning model. The paper also demonstrates the use of convolutional neural networks (CNNs) on raw IQ samples for modulation classification, achieving competitive accuracy compared to traditional feature-based methods. The authors discuss the potential benefits of DL in communications, including the ability to handle complex and non-linear channel impairments, and highlight open challenges and future research directions.The paper introduces several novel applications of deep learning (DL) in the physical layer of communication systems. By treating a communications system as an autoencoder, the authors propose a new approach to design end-to-end reconstruction tasks that jointly optimize transmitter and receiver components. This concept is extended to multiple transmitters and receivers, leading to the introduction of radio transformer networks (RTNs) which integrate domain-specific knowledge into the machine learning model. The paper also demonstrates the use of convolutional neural networks (CNNs) on raw IQ samples for modulation classification, achieving competitive accuracy compared to traditional feature-based methods. The authors discuss the potential benefits of DL in communications, including the ability to handle complex and non-linear channel impairments, and highlight open challenges and future research directions.