Could any one explain how the results on alphas_ attribute in Lars model are calculated? In the definition: alphas_ is the maximum covariance (abs value) in each iteration.
But when I look into details, I find that: The first element in alphas_ (in index=0) is the most important correlation (and not convariance) between features and the target. So, then we get our first feature. The second element in alphas_(index=1), is the first correlation found - first beta (for the first feature) But starting the third element, I couldn't found the right relationship between correlations and betas.
Could you please help?