|Thesis abstract: |
Iterative algorithms have found widespread use in many fields of the theory of modern information technology. The most known example is the implementation of message passing in low density parity-check (LDPC) decoding. The idea of a continue exchange of information between agents of a system is a powerful way of solving complex problems that are still either unsolved or that have a bad solution.
These methods can be applied in different field of interests; for example, the Belief Propagation method, used to marginalize the factor graph of a LDPC code, has been recently rearranged to solve trust and reputation problems. If we consider a scenario where we have raters that give an evaluation of rated performances, the problem can be modeled through a factor graph in the same way of LDPC codes. Hence, one can think to apply a method similar to Belief Propagation to work out which rater is a good rater and some other of his features; one can use this information to improve the confidence of the performances. Moreover, the algorithm can be extended to find out if a correlation exists between a group of raters and a group of rated performances, and this can help finding out if someone is cheating on his evaluations or not.
Another potentially innovative application of iterative algorithms concerns with their use in multiuser interference limited MIMO channels. In this context , the knowledge of the channel both at the transmitter and at the receiver allows one to focus on signal interactions in place of treating interference as noise. An increase capacity can be achieved when an accurate estimate of the channel is available. The goal is that of considering iterative techniques to achieve a good description of the channel while keeping the overhead low.