The focus of this paper is in Q-Lasso introduced in Alghamdi et al. (2013) which extended the Lasso by Tibshirani (1996). The closed convex subset Q belonging in a Euclidean m-space, for , is the set of errors when linear measurements are taken to recover a signal/image via the Lasso. Based on a recent work by Wang (2013), we are interested in two new penalty methods for Q-Lasso relying on two types of difference of convex functions (DC for short) programming where the DC objective functions are the difference of and norms and the difference of and norms with . By means of a generalized q-term shrinkage operator upon the special structure of norm, we design a proximal gradient algorithm for handling the DC model. Then, based on the majorization scheme, we develop a majorized penalty algorithm for the DC model. The convergence results of our new algorithms are presented as well. We would like to emphasize that extensive simulation results in the case show that these two new algorithms offer improved signal recovery performance and require reduced computational effort relative to state-of-the-art and () models, see Wang (2013). We also devise two DC Algorithms on the spirit of a paper where exact DC representation of the cardinality constraint is investigated and which also used the largest-q norm of and presented numerical results that show the efficiency of our DC Algorithm in comparison with other methods using other penalty terms in the context of quadratic programing, see Jun-ya et al. (2017).
1. Introduction and preliminaries
The process of compressive sensing (CS) [8], which consists of encoding and decoding, is rapidly consolidated year after year due to the blooming of large datasets which become increasingly important and available. The process of encoding involves taking a set of (linear) measurements, , where A is a matrix of size . If , we can compress the signal , whereas the process of decoding is to recover x from b where x is assumed to be sparse. It can be formulated as an optimization problem, namely
where is the norm, which counts the number of nonzero entries of x; namely
with being here the cardinality, i.e., the number of elements of a set. Hence minimizing the norm amounts to finding the sparsest solution. One of the difficulties in CS is solving the decoding problem above, since optimization is NP-hard. An approach that has gained popularity is to replace by the convex norm since it often gives a satisfactory sparse solution and has been applied in many different fields such as geology and ultrasound imaging.
More recently, nonconvex metrics were used as alternative approaches to , especially the nonconvex metric for in [6] which can be interpreted as a continued approximation strategy of as . A great deal of research has been conducted into problems including all kinds of variants and related algorithms, as you can see in [4] and references therein. The convex relaxation compared to the nonconvex problem () is generally more difficult to handle. However, it was shown in [12] that the potential reduction method can solve this special nonconvex problem in polynomial time with arbitrarily given accuracy.
Most recently, the majority of such sparsity inducing functions are unified as the notion of DC programming in [9], including log-sum, smoothly clipped absolute deviation and capped- penalty. Generally, DC programming problem can be solved through a primal–dual convex relaxations algorithm which is famous in the literature of DC Programming [11]. Other algorithms appeared as for solving application problems of DC programming in the area of finance and insurance, data analysis, machine learning as well as signal processing. However, as noted in [18], among the above mentioned DC programming approaches for sparse reconstruction, most of them are mainly preserving the separability properties of both and norms.
To begin with, let us recall that the lasso of Tibshirani [16] is given by the following minimization problem
A being an real matrix, and is a tuning parameter. The latter is nothing else than the basic pursuit (BP) of Chen et al. [7], namely
However, the constraint being inexact due to errors of measurements, the problem (1.4) can be reformulated as
where is the tolerance level of errors and p is often or . It is noticed in [1] that (1.5) can be rewritten as
in the case when , the closed ball in with center b and radius .
Now, when Q is a nonempty closed convex set of and the orthogonal projection from onto the set Q and by observing that the constraint is equivalent to the condition , this leads to the following Lagrangian formulation
being a Lagrangian multiplier.
A link is also made in [1] with split feasibility problems [5] which consist in finding x satisfying
with C and Q two nonempty closed convex subsets of and , respectively. An equivalent formulation of (1.8) as a minimization problem is given by
and its -regularization is
with a regularization parameter.
This convex relaxation approach was frequently employed, see for example [1,20] and references there in. As the level curves of - are closer to than those of , this motivated us in [14] to propose a regularization of split feasibility problems by means of the nonconvex -, namely
and present three algorithms with their convergence properties [14]. Unlike the separable sparsity inducing functions involved in the aforementioned DC programming for problem , we are interested in the two first sections of this work to two specific types of DC programming with un-separable objective functions, which are in the form of difference functions between two norms, namely the new notion denoting the sum of q largest elements of a vector in magnitude (i.e., the norm of q-term best approximation of a vector) introduced in [18] and the classical norm with . Obviously and are regular convex norms. The corresponding DC programs are as follows:
and
where is defined as the sum of the q largest elements of x in magnitude, and . We would like to emphasize that the following least-squares variant of (1.12) and (1.13), were studied in the recent work by Wang [18]:
where and , and
where and .
This paper proposes generalizations to Q-Lasso, namely
where and , as well as
where and , and our attention will be focused on the algorithmic aspect.
The rest of the paper is organized as follows. In Sections 2 and 3, two DC-penalty methods instead of conventional methods such as or minimization are proposed. Their convergence to a stationary point are also analyzed. The first iterative minimization method is based on the gradient proximal algorithm and the second one is designed by means of the majored penalty strategy. Furthermore, relying on DCA (difference of convex algorithm) two other algorithms are proposed and their convergence results are established in Section 3 and 4.
2. Proximal gradient algorithm
First, we recall that the subdifferential of a convex function is given by
Each element of is called subgradient. If , it is well-known that
and when , we have
The indicator function of a set is defined by
Moreover, the normal cone of a set C at , denoted by is defined as
Connection between the above definitions is given by the key relation .
In this section our interest is in solving the DC programming
where and .
Similar to norm, norm, etc., we adopt the notation to denote the norm of which is defined a line below (1.13) and we design an iterative algorithm based both on a generalized q-term shrinkage operator and on the proximal gradient algorithm framework.
At this stage, observe that the restriction on guarantees that for all x. To solve (2.6), we consider the following standard proximal gradient algorithm:
1. Initialization: Let be given and set with the maximal eigenvalue.
2. For find
Observe that subproblem (2.7) can equivalently formulated as
Thus, it suffices to consider the solutions to the following minimization problem
with a given vector y and positive numbers . An explicit solution of this problem is given by the following result, see [18].
Proposition 2.1 Let be the indices such that
Then with
is a solution of(2.9).
The proximal operator above (called the generalized q-term shrinkage operator in [18]) amounts to write the algorithm as follows:
Proximal Gradient Algorithm:
- 1.
Start: Let be given and set with the maximal eigenvalue.
- 2.
For find
Sort as
where .
End.
Now, we are in a position to show the following convergence result of the scheme (2.7):
Proposition 2.2 The sequence generated by the Proximal Gradient Algorithm above converges to a stationary point of problem (2.6).
Proof. Remember that is differentiable and its gradient is Lipschitz continuous with constant . By [3]-Proposition A.24, we have
Combining this with definition of , we obtain
Since , we see immediately that and thus the sequence is convergent since f is a non-negative function. Furthermore, we obtain that which follows by summing (2.12) from to . As a further consequence, we note that
Since , we have that is bounded. Moreover, the objective function f is square term plus a piecewise linear function which ensures that f is semi-algebric and hence satisfies Kurdyka-Lojasiewicz inequality. [2]-Theorem 5.1 is then applicable and obtain that is convergent to a stationary point of (2.6). □
3. Majorized penalty algorithm
Consider the following minimization problem
where a nonempty closed convex set of and .
First, observe again that conditions on guarantees that for all x. We will now describe an algorithm for solving (3.1), based on the majorized penalty approach see, for example, [18] and references therein. Following the same lines as in [18], we start by constructing a majorization of . To that end let , then for any , we have
Moreover, by invoking the convexity of the norm and definition of its subdifferential, we also have
where
Hence, if we define
hence, for every , we get
Starting with an initial iterate , the majorized penalty approach above updates by solving
This leads to the following explicit formulation of by means of the proximity (shrinkage) operator of :
where
We summarize the algorithm as follows:
Majorized Penalty Algorithm:
1. Initialization: Let be given and set .
2. For find
End.
The following proposition contains the convergence result of this Penalty Algorithm.
Proposition 3.1 Let be the sequence generated by the Majorized Penalty Algorithm above. Then
Furthermore, the sequence is bounded and any cluster point is a stationary point of problem (3.1).
Proof. Since minimizes , thanks to the first-order optimality condition we can write
being a subgradient of at . This combined with the definition of the subdifferential of at gives
Hence
This together with the definition of F, for any , leads to
Consequently,
Hence and thus the sequence is convergent since is a non-negative function. Furthermore, the sequence is such that
Indeed, by summing (3.7) from to , we obtain that
Consequently, the sequence is asymptotically regular, i.e., . On the other hand, observe that the definition of for any , leads to
Since , we obtain that . This implies that is bounded since . To conclude, we prove that every cluster point of is a stationary point of (3.1). Let be a cluster point of , then being subsequence of . By passing to the limit in (3.6) along the subsequence and in the light of the upper semicontinuity of (Clarke) subdifferentials, we obtain the desired result, namely
which is nothing else than the first-order optimality condition of (3.1). □
4. DCA algorithm
Now we turn our attention to a DC Algorithm (DCA), where the dual step at each iteration can be efficiently carried out due to the accessible subgradients of the largest- q-norm and norm. Remember that to find critical points of , the DCA consists in designing of sequences and by the following rules
Note that by the definition of subdifferential, we can write
Since minimizes , we also have
Combining the last inequalities, we obtain
Therefore, the DCA leads to a monotonically decreasing sequence that converges as long as the objective function f is bounded below.
Now, we can decompose the objective function in (2.6) as follows
where , , here and .
At each iteration, DCA solves The convex subproblem defined by linearizing the concave term is solved by DCA at each iteration until a convergence condition is satisfied. More precisely, we have
Especially, if either the function or is polyhedral, the DCA is said to be polyhedral and terminates in finite iterations [15]. Note that the our proposed DCA is polyhedral since the largest-q norm term can be expressed as a pointwise maximum of linear functions, see [10]. On the other hand, the subdifferential of at a point is given in, see for example [19],
that is
where denotes the element of y corresponding to in the linear program (4.4). Observe that a subgradient can be computed efficiently by first sorting the elements in decreasing order, namely . Then, assign 1 to which corresponds to .
To conclude, let us consider the following DC formulation of (3.1):
where , , here and .
The subgradient is also available via the formula (3.2) and the DCA in this context take the following form
where
For the details of DCA convergence properties, see [15].
5. Concluding remarks
The focus of this paper is on Q-Lasso relying on two new DC-penalty methods instead of conventional methods such as or minimization developed in [13,17] and [21]. Two iterative minimization methods based on the gradient proximal algorithm as well as the majored penalty algorithm are designed and their convergence to a stationary point is proved. Furthermore, by means of DC (difference of convex) Algorithm, two other algorithms are devised and their convergence results are also stated.
Publishers note: The publisher wishes to inform readers that the article “Difference of two norms-regularizations for Q-Lasso” was originally published by the previous publisher of Applied Computing and Informatics and the pagination of this article has been subsequently changed. There has been no change to the content of the article. This change was necessary for the journal to transition from the previous publisher to the new one. The publisher sincerely apologises for any inconvenience caused. To access and cite this article, please use Moudafi, A. (2020), “Difference of two norms-regularizations for Q-Lasso”, New England Journal of Entrepreneurship. Vol. 17 No. 1, pp. 79-89. The original publication date for this paper was 19/17/2018.
