Sample and Computationally Efficient Learning Algorithms under S-Concave Distributions