Graphical LASSO based Model Selection for Time Series

A Note on the Lasso for Gaussian Graphical Model Selection

Inspired by the success of the Lasso for regression analysis (Tibshirani, 1996), it seems attractive to estimate the graph of a multivariate normal distribution by `1-norm penalised likelihood maximisation. The objective function is convex and the graph estimator can thus be computed efficiently, even for very large graphs. However, we show in this note that the resulting estimator is not consi...

Decomposability and selection of graphical models for multivariate time series

SVD-Based Screening for the Graphical Lasso

The graphical lasso is the most popular approach to estimating the inverse covariance matrix of highdimension data. It iteratively estimates each row and column of the matrix in a round-robin style until convergence. However, the graphical lasso is infeasible due to its high computation cost for large size of datasets. This paper proposes Sting, a fast approach to the graphical lasso. In order ...

Metric-based model selection for time-series forecasting

Metric-based methods, which use unlabeled data to detect gross differences in behavior away from the training points, have recently been introduced for model selection, often yielding very significant improvements over alternatives (including cross-validation). We introduce extensions that take advantage of the particular case of time-series data in which the task involves prediction with a hor...

Graphical Interaction Models for Time Series: Parameter Estimation and Model Selection

We present a parametric approach for graphical interaction modelling in multivariate stationary time series. In these models, the possible dependencies between the components of the process are represented by edges in an undirected graph. We consider vector autoregressive models and propose a parametrization in terms of inverse covariances, which are contrained to zero for missing edges. The pa...