Secure multi - party computation and its application
Time: 2020-12-16 15:46:00
Topic：Secure multi - party computation and its application
Projection-type methods for solving variational inequalities usually require that the mapping involved have a certain kind of generalized monotonicity and/or be Lipschitz continuous, and compute projection once or twice to generate the next iteration from the current iteration. The original projection method computes one projection at each iteration, but needs very restrictive assumption that the mapping is strongly monotone and Lipschitz continuous. The assumption can be relaxed if the times of computing projection become twice instead of once, this is what the extragradient method and its variant do. This talk concerns a topic of growing interest which either computes projection twice at each iteration and assumes less restrictive monotonicity， or computes projection once at each iteration with only monotonicity assumed.