Sparsity-Cognizant Total Least-Squares for Perturbed Compressive Sampling
Citations
797 citations
623 citations
Cites background or methods from "Sparsity-Cognizant Total Least-Squa..."
...The off-grid distance (the distance from the true DOA to the nearest grid point) that lies in a bounded interval is assumed to be uniformly distributed (noninformative) rather than Gaussian as in [17]....
[...]
...In the following we re-derive the off-grid model proposed in [17] using linear approximation and further show its relationship with the on-grid one....
[...]
...As argued in [5] the SVD used in OGSBI-SVD can alleviate the sensitivity to the measurement noise in the MMV case that is not used in [17]....
[...]
...An off-grid model for DOA estimation is studied in [17] where the estimated DOAs are no longer constrained in the sampling grid set....
[...]
...In this paper, we studied the off-grid DOA estimation model firstly proposed in [17] for reducing the modeling error due to discretization of a continuous range....
[...]
473 citations
318 citations
Cites methods from "Sparsity-Cognizant Total Least-Squa..."
...In [66], the sparse total least square (S-TLS) method has been proposed to solve the following single-snapshot problem:...
[...]
...Despite the single-snapshot data at hand [66], the proposed approach was shown to overcome LASSO strategies [35]....
[...]
202 citations
References
40,785 citations
"Sparsity-Cognizant Total Least-Squa..." refers methods in this paper
...From a high-level view, the novel scheme comprises anouter iteration loop based on the bisection method [14], and aninner iteration loop that relies on a variant of the branch-and-bound (BB) method [1]....
[...]
19,261 citations
19,211 citations
"Sparsity-Cognizant Total Least-Squa..." refers background or methods in this paper
...The Lagrangian form of BP is also popular in statistics for fitting sparse linear regression models, using the so-termed least-absolute shrinkage and selection operator (Lasso); see e.g., [ 19 ], [29], and references thereof....
[...]
...(For large , the solution is driven toward the all-zero vector; whereas for small it tends to the LS solution.) This form of BP coincides with the Lasso approach developed for variable selection in linear regression problems [ 19 ], [29]....
[...]
...Lasso problem is known to admit a closed-form solution expressed in terms of a soft thresholding operator (see, e.g., [ 19 ])...
[...]
...From the plethora of available options to solve (17), it is worth mentioning two computationally efficient ones: the least-angle regression (LARS), and the coordinate descent (CD); see e.g., [ 19 ]....
[...]
...Remark 2 (Group Lasso and Matrix S-TLS): When groups of entries are ap rioriknown to be zero or nonzero (as a group), the -norm in (3) must be replaced by the sum of -norms, namely . The resulting group S-TLS estimate can be obtained using the group-Lasso solver [ 19 ]....
[...]
10,549 citations
9,950 citations
Additional excerpts
...…that [7, Sec 3.3.4]: using the solution̂xS−TLS of (6) for a given multiplierλ > 0 and lettingµ := ‖x̂S−TLS‖1, the pertinent constraint isX1(µ) := {x ∈ Rn : ‖x‖1 ≤ µ}; and the equivalent constrained minimization problem is [cf. (6)] x̂S−TLS := arg min x∈X1(µ) f(x) , f(x) := ‖y −Ax‖22 1 + ‖x‖22 ....
[...]