Lasso-type Penalty Research Articles

A uniform framework for the combination of penalties in generalized structured models

Penalized estimation has become an established tool for regularization and model selection in regression models. A variety of penalties with specific features are available and effective algorithms for specific penalties have been proposed. But not much is available to fit models with a combination of different penalties. When modeling the rent data of Munich as in our application, various types of predictors call for a combination of a Ridge, a group Lasso and a Lasso-type penalty within one model. We propose to approximate penalties that are (semi-)norms of scalar linear transformations of the coefficient vector in generalized structured models--such that penalties of various kinds can be combined in one model. The approach is very general such that the Lasso, the fused Lasso, the Ridge, the smoothly clipped absolute deviation penalty, the elastic net and many more penalties are embedded. The computation is based on conventional penalized iteratively re-weighted least squares algorithms and hence, easy to implement. New penalties can be incorporated quickly. The approach is extended to penalties with vector based arguments. There are several possibilities to choose the penalty parameter(s). A software implementation is available. Some illustrative examples show promising results.

Selection and Fusion of Categorical Predictors with L0-Type Penalties

In regression modelling, categorical covariates have to be coded. Depending on the number of categorical covariates and on the number of levels they have, the number of coefficients can become huge. To reduce the model complexity, coefficients of similar categories should be fused and coefficients of non-influential categories should be set to zero. To this end, Lasso-type penalties on the differences of coefficients are a standard approach. However, the clustering/selection performance of this approach is sometimes poor–especially when the adaptive weights are badly conditioned or not existing. In some situations, there is no incentive to cluster similar categories. To overcome this, a [Formula: see text] penalty on the differences of coefficients is proposed, whereby the [Formula: see text] ‘norm’ is defined as the number of non-zero entries in a vector. The proposed penalty favours to find clusters of categories that share the same effect on the response variable while the estimation accuracy is comparable to Lasso-type penalties. Numerical experiments within the framework of generalized linear models are promising. For illustration, data on the unemployment rates in Germany is analyzed.

Using Lasso-Type Penalties to Model Time-Varying Covariate Effects in Panel Data Regressions A Novel Approach Illustrated by the 'Death of Distance' In International Trade

When analyzing panel data using regression models, it is often reasonable to allow for time-varying covariate eects. We propose a novel approach to modelling timevarying coecients in panel data regressions, which is based on penalized regression techniques. To illustrate the usefulness of this approach, we revisit the well-known empirical puzzle of the ‘death of distance’ in international trade. We find significant dierences between results obtained with the proposed estimator and those obtained with ‘traditional’ methods. The proposed method can also be used for model selection, and to allow covariate eects to vary over other dimensions than time.

Pattern alternating maximization algorithm for missing data in high-dimensional problems

We propose a novel and efficient algorithm for maximizing the observed log-likelihood of a multivariate normal data matrix with missing values. We show that our procedure, based on iteratively regressing the missing on the observed variables, generalizes the standard EM algorithm by alternating between different complete data spaces and performing the E-Step incrementally. In this non-standard setup we prove numerical convergence to a stationary point of the observed log-likelihood. For high-dimensional data, where the number of variables may greatly exceed sample size, we perform regularization using a Lasso-type penalty. This introduces sparsity in the regression coefficients used for imputation, permits fast computation and warrants competitive performance in terms of estimating the missing entries. We show on simulated and real data that the new method often improves upon other modern imputation techniques such as k-nearest neighbors imputation, nuclear norm minimization or a penalized likelihood approach with an l1-penalty on the concentration matrix.

Variable selection in high-dimensional quantile varying coefficient models

In this paper, we propose a two-stage variable selection procedure for high dimensional quantile varying coefficient models. The proposed method is based on basis function approximation and LASSO-type penalties. We show that the first stage penalized estimator with LASSO penalty reduces the model from ultra-high dimensional to a model that has size close to the true model, but contains the true model as a valid sub model. By applying adaptive LASSO penalty to the reduced model, the second stage excludes the remained irrelevant covariates, leading to an estimator consistent in variable selection. A simulation study and the analysis of a real data demonstrate that the proposed method performs quite well in finite samples, with regard to dimension reduction and variable selection.

Multivariate regression shrinkage and selection by canonical correlation analysis

The problem of regression shrinkage and selection for multivariate regression is considered. The goal is to consistently identify those variables relevant for regression. This is done not only for predictors but also for responses. To this end, a novel relationship between multivariate regression and canonical correlation is discovered. Subsequently, its equivalent least squares type formulation is constructed, and then the well developed adaptive LASSO type penalty and also a novel BIC-type selection criterion can be directly applied. Theoretical results show that the resulting estimator is selection consistent for not only predictors but also responses. Numerical studies are presented to corroborate our theoretical findings.

Quantile regression in high-dimension with breaking

The paper considers a linear regression model in high-dimension for which the predictive variables can change the influence on the response variable at unknown times (called changepoints). Moreover, the particular case of the heavy-tailed errors is considered. In this case, least square method with LASSO or adaptive LASSO penalty can not be used since the theoretical assumptions do not occur or the estimators are not robust. Then, the quantile model with SCAD penalty or median regression with LASSO-type penalty allows, in the same time, to estimate the parameters on every segment and eliminate the irrelevant variables. We show that, for the two penalized estimation methods, the oracle properties is not affected by the change-point estimation. Convergence rates of the estimators for the change-points and for the regression parameters, by the two methods are found. Monte-Carlo simulations illustrate the performance of the methods.

Structured kernel quantile regression

Quantile regression can provide more useful information on the conditional distribution of a response variable given covariates while classical regression provides informations on the conditional mean alone. In this paper, we propose a structured quantile estimation methodology in a nonparametric function estimation setup. Through the functional analysis of variance decomposition, the optimization of the proposed method can be solved using a series of quadratic and linear programmings. Our method automatically selects relevant covariates by adopting a lasso-type penalty. The performance of the proposed methodology is illustrated through numerical examples on both simulated and real data.

Multivariate Regression Shrinkage and Selection by Canonical Correlation Analysis

The problem of regression shrinkage and selection for multivariate regression is considered. The goal is to consistently identify those variables relevant for regression. This is done not only for predictors but also for responses. To this end, a novel relationship between multivariate regression and canonical correlation is discovered. Subsequently, its equivalent least squares type formulation is constructed, and then the well developed adaptive LASSO type penalty and also a novel BIC-type selection criterion can be directly applied. Theoretical results show that the resulting estimator is selection consistent for not only predictors but also responses. Numerical studies are presented to corroborate our theoretical findings.

Positive definite estimators of large covariance matrices

Using convex optimization, we construct a sparse estimator of the covariance matrix that is positive definite and performs well in high-dimensional settings. A lasso-type penalty is used to encourage sparsity and a logarithmic barrier function is used to enforce positive definiteness. Consistency and convergence rate bounds are established as both the number of variables and sample size diverge. An efficient computational algorithm is developed and the merits of the approach are illustrated with simulations and a speech signal classification example. Copyright 2012, Oxford University Press.

Kernel based methods for accelerated failure time model with ultra-high dimensional data

BackgroundMost genomic data have ultra-high dimensions with more than 10,000 genes (probes). Regularization methods with L1 and Lp penalty have been extensively studied in survival analysis with high-dimensional genomic data. However, when the sample size n ≪ m (the number of genes), directly identifying a small subset of genes from ultra-high (m > 10, 000) dimensional data is time-consuming and not computationally efficient. In current microarray analysis, what people really do is select a couple of thousands (or hundreds) of genes using univariate analysis or statistical tests, and then apply the LASSO-type penalty to further reduce the number of disease associated genes. This two-step procedure may introduce bias and inaccuracy and lead us to miss biologically important genes.ResultsThe accelerated failure time (AFT) model is a linear regression model and a useful alternative to the Cox model for survival analysis. In this paper, we propose a nonlinear kernel based AFT model and an efficient variable selection method with adaptive kernel ridge regression. Our proposed variable selection method is based on the kernel matrix and dual problem with a much smaller n × n matrix. It is very efficient when the number of unknown variables (genes) is much larger than the number of samples. Moreover, the primal variables are explicitly updated and the sparsity in the solution is exploited.ConclusionsOur proposed methods can simultaneously identify survival associated prognostic factors and predict survival outcomes with ultra-high dimensional genomic data. We have demonstrated the performance of our methods with both simulation and real data. The proposed method performs superbly with limited computational studies.

Statistical approaches to combining binary classifiers for multi-class classification

One of the popular methods for multi-class classification is to combine binary classifiers. In this paper, we propose a new approach for combining binary classifiers. Our method trains a combining method of binary classifiers using statistical techniques such as penalized logistic regression, stacking, and a sparsity promoting penalty. Our approach has several advantages. Firstly, our method outperforms existing methods even if the base classifiers are well-tuned. Secondly, an estimate of conditional probability for each class can be naturally obtained. Furthermore, we propose selecting relevant binary classifiers by adding the group lasso type penalty in training the combining method.

A Framework for Feature Selection in Clustering

We consider the problem of clustering observations using a potentially large set of features. One might expect that the true underlying clusters present in the data differ only with respect to a small fraction of the features, and will be missed if one clusters the observations using the full set of features. We propose a novel framework for sparse clustering, in which one clusters the observations using an adaptively chosen subset of the features. The method uses a lasso-type penalty to select the features. We use this framework to develop simple methods for sparse K-means and sparse hierarchical clustering. A single criterion governs both the selection of the features and the resulting clusters. These approaches are demonstrated on simulated and genomic data.

REMAS: a new regression model to identify alternative splicing events from exon array data

BackgroundAlternative splicing (AS) is an important regulatory mechanism for gene expression and protein diversity in eukaryotes. Previous studies have demonstrated that it can be causative for, or specific to splicing-related diseases. Understanding the regulation of AS will be helpful for diagnostic efforts and drug discoveries on those splicing-related diseases. As a novel exon-centric microarray platform, exon array enables a comprehensive analysis of AS by investigating the expression of known and predicted exons. Identifying of AS events from exon array has raised much attention, however, new and powerful algorithms for exon array data analysis are still absent till now.ResultsHere, we considered identifying of AS events in the framework of variable selection and developed a regression method for AS detection (REMAS). Firstly, features of alternatively spliced exons were scaled by reasonably defined variables. Secondly, we designed a hierarchical model which can represent gene structure and transcriptional influence to exons, and the lasso type penalties were introduced in calculation because of huge variable size. Thirdly, an iterative two-step algorithm was developed to select alternatively spliced genes and exons. To avoid negative effects introduced by small sample size, we ranked genes as parameters indicating their AS capabilities in an iterative manner. After that, both simulation and real data evaluation showed that REMAS could efficiently identify potential AS events, some of which had been validated by RT-PCR or supported by literature evidence.ConclusionAs a new lasso regression algorithm based on hierarchical model, REMAS has been demonstrated as a reliable and effective method to identify AS events from exon array data.

Simultaneous estimation and variable selection in median regression using Lasso-type penalty

We consider the median regression with a LASSO-type penalty term for variable selection. With the fixed number of variables in regression model, a two-stage method is proposed for simultaneous estimation and variable selection where the degree of penalty is adaptively chosen. A Bayesian information criterion type approach is proposed and used to obtain a data-driven procedure which is proved to automatically select asymptotically optimal tuning parameters. It is shown that the resultant estimator achieves the so-called oracle property. The combination of the median regression and LASSO penalty is computationally easy to implement via the standard linear programming. A random perturbation scheme can be made use of to get simple estimator of the standard error. Simulation studies are conducted to assess the finite-sample performance of the proposed method. We illustrate the methodology with a real example.

Sparse permutation invariant covariance estimation

The paper proposes a method for constructing a sparse estimator for the inverse covariance (concentration) matrix in high-dimensional settings. The estimator uses a penalized normal likelihood approach and forces sparsity by using a lasso-type penalty. We establish a rate of convergence in the Frobenius norm as both data dimension p and sample size n are allowed to grow, and show that the rate depends explicitly on how sparse the true concentration matrix is. We also show that a correlation-based version of the method exhibits better rates in the operator norm. We also derive a fast iterative algorithm for computing the estimator, which relies on the popular Cholesky decomposition of the inverse but produces a permutation-invariant estimator. The method is compared to other estimators on simulated data and on a real data example of tumor tissue classification using gene expression data.

Structured variable selection in support vector machines

When applying the support vector machine (SVM) to high-dimensional classification problems, we often impose a sparse structure in the SVM to eliminate the influences of the irrelevant predictors. The lasso and other variable selection techniques have been successfully used in the SVM to perform automatic variable selection. In some problems, there is a natural hierarchical structure among the variables. Thus, in order to have an interpretable SVM classifier, it is important to respect the heredity principle when enforcing the sparsity in the SVM. Many variable selection methods, however, do not respect the heredity principle. In this paper we enforce both sparsity and the heredity principle in the SVM by using the so-called structured variable selection (SVS) framework originally proposed in [20]. We minimize the empirical hinge loss under a set of linear inequality constraints and a lasso-type penalty. The solution always obeys the desired heredity principle and enjoys sparsity. The new SVM classifier can be efficiently fitted, because the optimization problem is a linear program. Another contribution of this work is to present a nonparametric extension of the SVS framework, and we propose nonparametric heredity SVMs. Simulated and real data are used to illustrate the merits of the proposed method.

Lasso type classifiers with a reject option

We consider the problem of binary classification where one can, for a particular cost, choose not to classify an observation. We present a simple proof for the oracle inequality for the excess risk of structural risk minimizers using a lasso type penalty.