When can you trust feature selection? -- I: A condition-based analysis of LASSO and generalised hardness of approximation

Open in new window