This paper presents the implementation of two methods for sparse index tracking of the US Nasdaq 100 index. The proposed methods employ a regularization approach based on a non-convex cardinality constraint approximation of the ℓp-norm to identify the optimal asset weights of the tracking portfolio. The results demonstrate that the ℓ0-norm-constrained sparse tracking portfolio is computationally efficient and exhibits a notable reduction in tracking errors during the out-of-sample testing periods. Furthermore, we present an empirical comparison of the results of performance measures with those of traditional constrained strategies using norm constraints for index tracking.