Submodular Secretary Problem with Shortlists under General Constraints

In submodular k-secretary problem, the goal is to select k items in a randomly ordered input so as to maximize the expected value of a given monotone submodular function on the set of selected items. In this paper, we introduce a relaxation of this problem, which we refer to as submodular k-secretary problem with shortlists. In the proposed problem setting, the algorithm is allowed to choose more than k items as part of a shortlist. Then, after seeing the entire input, the algorithm can choose a subset of size k from the bigger set of items in the shortlist. We are interested in understanding to what extent this relaxation can improve the achievable competitive ratio for the submodular k-secretary problem. In particular, using an O(k) shortlist, can an online algorithm achieve a competitive ratio close to the best achievable online approximation factor for this problem? We answer this question affirmatively by giving a polynomial time algorithm that achieves a 1 - 1/e - epsilon -O(k^{-1}) competitive ratio for any constant epsilon>0, using a shortlist of size eta {epsilon}(k)=O(k). Also, for the special case of m-submodular functions, we demonstrate an algorithm that achieves a 1 - epsilon competitive ratio for any constant epsilon > 0, using an O(1) shortlist. Finally, we show that our algorithm can be implemented in the streaming setting using a memory buffer of size eta{epsilon}(k)=O(k) to achieve a 1 - 1/e - epsilon - O(k^{-1}) approximation for submodular function maximization in the random order streaming model. This substantially improves upon the previously best known approximation factor of 1/2 + 8*10^{-14} [Norouzi-Fard et al. 2018] that used a memory buffer of size O(k log k). We further generalize our results to the case of matroid constraints. We design an algorithm that achieves a 1/2(1 - 1/e^2 - epsilon - O(1/k)) competitive ratio for any constant epsilon>0, using a shortlist of size O(k). This is especially surprising considering that the best known competitive ratio for the matroid secretary problem is O(log log k). An important application of our algorithm is for the random order streaming of submodular functions. We show that our algorithm can be implemented in the streaming setting using O(k) memory. It achieves a 1/2 (1 - 1/e^2 - epsilon - O(1/k)) approximation. The previously best known approximation ratio for streaming submodular maximization under matroid constraint is 0.25 (adversarial order) due to [Feldman et al.], [Chekuri et al.], and [Chakrabarti et al.]. Moreover, we generalize our results to the case of p-matchoid constraints and give a frac{1}{p+1}(1 - 1/e^{p+1} - epsilon - O(1/k)) approximation using O(k) memory, which asymptotically approaches the best known offline guarantee frac{1}{p+1} [Nemhauser et al.]. Finally we empirically evaluate our results on real world data sets such as YouTube video and Twitter stream.