deflation Methods for Sparse PCA
- 1 Introduction
- 2 Deflation methods
- 3 Reformulating sparse PCA
- 4 Experiment
- 5 Conclusion
- 6 Relationship to Other Sparse PCA Methods
- 7 Reference
Principal component analysis (PCA) is a popular change of variables technique used in dimension reduction and visualization. The goal of PCA is to extract several principal components, linear combinations of input variables that together best account for the variance in a data set. Often, PCA is formulated as an eigenvalue decomposition problem: each eigenvector of the sample covariance matrix of a data set corresponds to the loadings or coefficients of a principal component. A common approach to solving this partial eigenvalue decomposition is to iteratively alternate between two sub-problems: rank-one variance maximization (which is done by finding the leading eigenvector of the covariance matrix) and matrix deflation (which is done by modifying the covariance matrix to eliminate the influence of that eigenvector).
A primary drawback of PCA is its lack of sparsity. Each principal component is a linear combination of all variables; the loadings (weights) are all typically non-zero. Sparsity is desirable as it often leads to more interpretable results, reduced computation time, and improved generalization. One approach to avoid this drawback of PCA is the formulation of sparse PCA (SPCA), a link to the 2004 paper of which by Iain M. Johnstone et al.<ref name="R1">Iain M. Johnstone and Arthur Y. Lu. "Sparse Principal Components Analysis". Technical report, Stanford University, Dept. of Statistics, 2004.</ref> is available here. SPCA makes the PCA process sparse by searching for pseudo-eigenvectors and their corresponding sparse (few such values are non-zero) loadings that explain the maximal amount of variance in the data.
In analogy to the PCA setting, many authors attempt to solve the sparse PCA problem by iteratively alternating between two sub-tasks: cardinality-constrained rank-one variance maximization and matrix deflation. The first sub-task is an NP-hard problem, and a variety of relaxations and approximate solutions have been developed in the literature. The latter sub-task has received relatively little attention and is typically borrowed without justification from the PCA context. In this paper, the author demonstrated that the standard PCA deflation procedure is seldom appropriate for the sparse PCA setting. To rectify the situation, the author first develops several heuristic deflation alternatives with more desirable properties, then reformulates the sparse PCA optimization problem to explicitly reflect the maximum additional variance objective on each round. The result is a generalized deflation procedure that typically outperforms more standard techniques on real-world data-sets.
is the identity matrix. is the set of all symmetric, positive semidefinite matrices in . represents the cardinality of or the number of non-zero entries in the vector .
A matrix deflation modifies a matrix to eliminate the influence of a given eigenvector, typically by setting the associated eigenvalue to zero.
Hotelling's deflation and PCA
In the PCA setting, the goal is to extract the
and we then use Hotelling's deflation to annihilate
The deflation step ensures that the-st leading eigenvector of is the leading eigenvector of . The following proposition explains why.
Ifare the eigenvalues of are the corresponding eigenvectors, and for some then has eigenvectors with corresponding eigenvalues
The proof of this proposition is simple and is given as follows:
, . QED
Thus, Hotelling’s deflation preserves all eigenvectors ofwhilst removing a selected eigenvalue of . This implies that Hotelling's deflation preserves the positive semi-definiteness of , i.e. the matrix resulting from deflating is positive semi-definite if and only if is positive semi-definite.
For the interested reader, a paper by Wang et al. titled Sparse PCA by Iterative Elimination Algorithm provides the details regarding this algorithm and is available here.
Hotelling's deflation and sparse PCA
In the sparse PCA setting, we seek sparse loadings which together capture the maximum amount of variance in the data. To find the first such "pseudo-eigenvector" equation, we can consider a cardinality-constrained version of the problem:
Typically, Hotelling’s deflation is utilized by substituting an extracted pseudo-eigenvector for a true eigenvector in the deflation step of Cauchy-Schwarz inequality is violated and the pseudo-eigenvectors may not be orthogonal to the deflated matrix. As well, the deflated matrix may not be positive semidefinite under pseudo-eigenvector Hotelling's deflation. This shortcoming can be a serious problem since most iterative sparse PCA methods assume a positive semidefinite matrix at each step.. This substitution, however, is seldom justified, for the properties of Hotelling’s deflation depend crucially on the use of a true eigenvector. If pseudo-eigenvectors are used in place of true eigenvectors, the
Alternative deflation techniques
To rectify the failings of pseudo-eigenvector Hotelling's deflation the author considers several alternative deflation techniques better suited to the sparse PCA setting. It should be noted that any deflation-based sparse PCA method can utilize any of the deflation techniques discussed below.
Given a data matrixand an arbitrary unit vector , an intuitive way to remove the contribution of from is to project onto the orthocomplement of the space spanned by . If is the sample covariance matrix of , then the sample covariance of is given by , which leads to our formulation for projection deflation:
- When is a true eigenvector of with eigenvalue , projection deflation reduces to Hotelling's deflation. The derivation of this property is shown below.
- In the general case in which is not a true eigenvector, projection deflation maintains the desirable properties that were lost to Hotelling's deflation. Unlike Hotelling's deflation, projection deflation preserves positive semi-definiteness of a matrix. This property is easy to show, and its derivation is given as follows: , where Also, unlike Hotelling's deflation, projection deflation renders both left and right orthogonal to . This is because and is symmetric. This property also implies another important property of projection deflation: projection deflation annihilates all covariances with .
Schur complement deflation
Let Schur complement of the vector variance in the full covariance matrix . Substituting sample covariance matrices for their population counterparts, Schur complement deflation is:be a unit vector and be a Gaussian random vector, representing the joint distribution of the data variables. If has covariance matrix , then ( ) has covariance matrix , and whenever . That is, the conditional variance is the
Like projection deflation, Schur complement deflation preserves positive-semi-definiteness and also rendersboth left and right orthogonal to . These two properties of Schur complement deflation are easy to show, and their derivations can be found in the Reference section of <ref name="R2">Lester Mackey. "Deflation Methods for Sparse PCA." In Advances in Neural Information Processing Systems 21, pages 1017–1024, 2009.</ref>
- When is an eigenvector of with eigenvalue , it reduces to Hotelling's deflation.
It should be noted that, even though we motivated Schur complement deflation with a Gaussianity assumption, in general, this technique has a more general interpretation as a column projection of a data matrix.
Whenever we deal with a sequence of non-orthogonal vectors, we must take care to distinguish between the variance explained by a vector and the additional variance explained, given all previous vectors. The additional variance explained by the-th pseudo-eigenvector, , is equivalent to the variance explained by the component of orthogonal to the space spanned by all previous pseudo-eigenvectors, where is the orthogonal projection onto the space spanned by . On each deflation step, therefore, we only want to eliminate the variance associated with . This is because annihilating the full vector would often lead to double counting by re-introducing components that are parallel to previously annihilated vectors. We shall now look at how this issue can be resolved. First, we can consider the issue of non-orthogonality in the context of Hotelling’s deflation. A modified version of Hotelling's deflation as described in  of the paper in the Reference section, known as orthogonalized Hotelling’s deflation (OHD), has been introduced. For this deflation technique, the following hold true:
- OHD is equivalent to Hotelling's deflation for :
- When Gram-Schmidt decomposition: , OHD can be expressed in terms of running
where , and form the columns of . Since form an orthonormal basis for the space spanned by , we have , which is the aforementioned orthogonal projection.
Because the first round of OHD is equivalent to an application of Hotelling’s deflation, OHD inherits all of the weaknesses associated with Hotelling's deflation. Fortunately, the same principles behind OHD that preserve orthogonality may also be applied to projection deflation to create an orthogonalized variant of projection deflation that inherits all of the desirable properties of projection deflation that Hotelling's deflation lacks. Furthermore, it should be noted that the approach described above is unnecessary for Schur complement deflation. This is because Schur complement deflation is unique in that it preserves orthogonality in all subsequent rounds of its application.
Proposition: Orthogonalized Schur complement deflation is equivalent to Schur complement deflation.
Proof: Consider the -th round of Schur complement deflation. We may write , where is in the subspace spanned by all previously extracted pseudo-eigenvectors and is orthogonal to this subspace. Then we know that , as is a linear combination of , and Thus, . Further, . Hence as
Summary of sparse PCA deflation method properties
The following table compares properties of the various deflation techniques studied above.
Reformulating sparse PCA
Recall that the goal of sparse PCA is to find cardinality-constrained pseudo-eigenvectors which together explain the most variance in the data. If we additionally constrain the sparse loadings to be generated sequentially, as in the PCA setting and the previous section, then a greedy approach of maximizing the additional variance of each new vector naturally suggests itself. On round , the additional variance of a vector is given by , where is the data covariance matrix, , and is the projection onto the space spanned by the previous pseudo-eigenvectors . As , maximizing the additional variance is equivalent to solving a cardinality-constrained maximum generalized eigenvalue problem,
If we let, then form an orthonormal basis for the space spanned by . Writing suggests a generalized deflation techniques that leads to the solution on each round.
Generalized deflation method for sparse PCA: Given ,
Gene expression data
The Berkeley Drosophila Transcription Network Project (BDTNP) 3D gene expression data contains gene expression levels measured in each nucleus of developing Drosophila embryos and averaged across many embryos and developmental stages. An aggregate VirtualEmbryo contains 21 genes and 5759 example nuclei. With this data we can test the various deflation techniques. Using the GSLDA algorithm <ref name="R3">B.Moghaddam, Y. Weiss, and S. Avidan. "Generalized spectral bounds for sparse LDA." In Proc. ICML, 2006.</ref> to solve the cardinality-constrained rank-one variance maximization and the following deflation techniques: Hotelling's (HD), projection )PD), Schur complement (SCD), orthogonalized Hotelling's (OHD), orthogonalized projection (OPD) and generalized(GD) we get the following results.
It shows a clear hierarchy among the deflation methods and the generalized deflation method performs best.
The utility of above Hotelling's deflation procedures is not limited to the sparse PCA setting. Indeed, the methods presented can be applied to any of a number of constrained eigendecomposition-based problems, including sparse canonical correlation analysis and linear discriminant analysis.
Relationship to Other Sparse PCA Methods