LIDA project overview
The focus of the project is on the joint research in the area of information theory, artificial intelligence and application of learning algorithms in information communication systems.
This project is the next step in cooperation between University of Belgrade and University of Arizona, and it results in joint research, invited lectures and a close interaction among PhD students from the two research groups. Two research groups plan joint applications for further project grants.
The main objective of the project is to develop new iterative decoding algorithms that could enable a significant increase of data rate in information communication systems. Possible applications of the algorithms is in energy efficient cellular systems (5G and beyond), satellite communications systems, quantum communications and computing and data storage.
The first specific objective of this research is to develop a principled deep neural network (DNN) framework for encoder/decoder design using the knowledge of the noisy channel subject to complexity and performance constraints. The second specific objective is the optimization of the gradient bit flipping decoder. The probabilistic variant of this decoder is proposed nine years ago by the project team members, and it is referenced in more than 80 journal and conference papers. Using the contemporary optimization methods, related to machine learning, in this project we have further improved the performance of the GDBF decoder.