Advances in IoT for Federated Machine Learning With GAN Applied to Gesture Recognition

Among the broad and diverse advances and efforts on distributed computing paradigm related research, such as Cloud Computing and Fog/Edge Computing, can be highlighted the Internet of Things (IoT) strengthening. IoT paradigm currently encompasses very diverse fields of application, including its use in Medical Applications, Medical IoT (MIoT), Industrial IoT (IIoT), as well as in developments related to Wearable IoT (WIoT) and the immersive experiences in the entertainment industry. At the same time, promising results related to Federated Machine Learning paradigm (FML), in particular using Generative Adversarial Networks (GANs), can be found in recently presented works. These FML strategies propose the use of computational resources close to the user, minimizing communication time to Cloud Data centers, like Edge Computing, using both Distributed Discriminants and Distributed Generators / Distributed Discriminants techniques. Based on our own experience, these strategies can be expected to be suitable for IoT processing. In that sense, by asking ourselves which of them could be most useful for Gesture Recognition and Facial Expression Recognition, advances was made with the analysis of several ideas in the state of the art. Server based and distributed solutions were analysed, considering both their theoretical robustness and their ability to adapt to distributed technologies or Federated Machine Learning strategies. This work brings the first stage of project development, in which the experimental domain was delimited, as well as the simulation and prototyping times needed to find useful solutions with good trade-off between computing power and sustainable energy consumption.