Strongly Truthful Interactive Regret Minimization

Being Happy with the Least: Achieving α-happiness with Minimum Number of Tuples

“…Greedy-based algorithms which construct the solutions iteratively were proposed in [9], [15], [16], [26]. Besides, the k-regret query can be solved in a set-cover manner [4] while user interactions were considered in [14], [24], [25].…”

Interactive Search for One of the Top-k

“…Assume that Alice would like to buy a cheap used car with high horse power. In literature [22,27,36], Alice's preference is represented by a monotonic function, called a utility function. Based on the function, each car has a utility (i.e., the function score).…”

“…the user's utility function. It assumes that a user knows his/her utility function precisely [22,36]. However, in practice, it is very likely that Alice has difficulty in specifying that her utility function has weight 41.2% for price and 58.8% for horse power.…”

“…Based on the user's feedback during interaction, the user's utility function is implicitly learned and one of the top-tuples is guaranteed to be returned. To interact with the user, following [27,36], for each question, we present the user with two candidate tuples and ask the user which tuple s/he prefers. This kind of interaction naturally appears in our daily life.…”

“…To the best of our knowledge, we are the first to study problem IST. Some closely related problems are [14,22,27,36] but they are different from ours. [14] also involves user interaction but it wants to find the ranking of all tuples (including the top-tuples) by interacting with the user.…”