Bolasso

The purpose of this Git is to resume the results of the article [Bach, F. R. (2008)]

This article is part of the analysis of parsimonious signals, in connection with our course on parsimony at INSA Rennes.

This article proposes new results on Lasso type problems by proposing a method based on resampling in order to solve the inconsistency problems of the Lasso regression.

Francis R. Bach proposes to use the bootstrap method and an ensemblistic philosophy (combining models on each new sample).

In this Git you will find 3 folders:

• utils: this folder contains an implementation of the Bolasso algorithm as well as the synthetic data generation algorithm used in the article
• src: this folder contains the realization of the experiments
• fig: this folder contains the figures of the experiments

Citations

Bach, F. R. (2008). Bolasso : Model consistent Lasso estimation through the bootstrap. In Proceedings of the 25th International Conference on Machine Learning - ICML ’08, pages 33–40, Helsinki, Finland. ACM Press.