TY - GEN
T1 - Dynamics of Damped Approximate Message Passing Algorithms
AU - Mimura, Kazushi
AU - Takeuchi, Jun'Ichi
PY - 2019/8
Y1 - 2019/8
N2 - For linear system models, the approximate massage passing (AMP) is one of the effective iterative sparse recovery algorithms. However, depending on a measurement matrix ensemble, AMP may face convergence issues. Some algorithms are proposed so far to avoid the convergence issues, e.g., the orthogonal AMP (OAMP) and the mean removal. One of the simplest ways to avoid the convergence issues is to introduce a damping effect into AMP. In this paper, we derive a simple recursive equations that characterizes the damped OAMP, which is an OAMP in which the damping effect is introduced, and show that the result can be applied to the damped version of the original AMP.
AB - For linear system models, the approximate massage passing (AMP) is one of the effective iterative sparse recovery algorithms. However, depending on a measurement matrix ensemble, AMP may face convergence issues. Some algorithms are proposed so far to avoid the convergence issues, e.g., the orthogonal AMP (OAMP) and the mean removal. One of the simplest ways to avoid the convergence issues is to introduce a damping effect into AMP. In this paper, we derive a simple recursive equations that characterizes the damped OAMP, which is an OAMP in which the damping effect is introduced, and show that the result can be applied to the damped version of the original AMP.
UR - http://www.scopus.com/inward/record.url?scp=85081100937&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85081100937&partnerID=8YFLogxK
U2 - 10.1109/ITW44776.2019.8989237
DO - 10.1109/ITW44776.2019.8989237
M3 - Conference contribution
T3 - 2019 IEEE Information Theory Workshop, ITW 2019
BT - 2019 IEEE Information Theory Workshop, ITW 2019
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2019 IEEE Information Theory Workshop, ITW 2019
Y2 - 25 August 2019 through 28 August 2019
ER -