Abstract We consider estimating a piecewise-constant image, or gradient-sparse signal on general graph, from noisy linear measurements. propose and study an iterative algorithm to minimize penalized least-squares objective, with penalty given by the “ℓ0-norm” of signal’s discrete graph gradient. The method uses non-convex variant proximal gradient descent, applying alpha-expansion procedure app...

A recent research direction in data-driven modeling is the identification of dynamic networks, which measured vertex signals are interconnected by edges represented causal linear transfer functions. The major question addressed this paper where to allocate external excitation such that a network model set becomes generically identifiable when measuring all signals. To tackle synthesis problem, ...