Convergence Analysis
In [35], it has been proven that for a Hermitian matrix,
all the components have independent differentials, and formal
derivatives/gradients should be used in optimization problems
with complex-valued matrix variables. Real scalars can be
viewed as one-dimensional Hermitian matrices and the above
results also apply. Following the mathematical results in [35],
for q = argminqL(q,',w), by the first-order optimality
condition, there exists a subgradient ∇f(q) [27], satisfying
Convergence AnalysisIn [35], it has been proven that for a Hermitian matrix,all the components have independent differentials, and formalderivatives/gradients should be used in optimization problemswith complex-valued matrix variables. Real scalars can beviewed as one-dimensional Hermitian matrices and the aboveresults also apply. Following the mathematical results in [35],for q = argminqL(q,',w), by the first-order optimalitycondition, there exists a subgradient ∇f(q) [27], satisfying
การแปล กรุณารอสักครู่..