Fig. 3From: Structured feature selection using coordinate descent optimization ℓ 1 and ℓ ∞ norms. ℓ ∞ is an upper bound for ℓ 1 normBack to article page