A-MCTS: Adaptive Monte Carlo Tree Search for Temporal Path Discovery

An attributed dynamic graph contains multiple dynamic attributes associated with each edge in the graph. In attributed dynamic graph based applications, people usually can specify multiple constraints in the attributes to illustrate their requirements, such as the total cost, the total travel time and the stopover interval of a flight between two cities. This inspires the multi-constrained temporal path discovery in attributed dynamic graphs, which is a challenging NP-complete problem. To solve the problem, the existing methods adopt reinforcement learning with Monte Carlo tree search. However, these reinforcement learning based methods require a certain degree of “discovery experience” to obtain better results, which can lead to the expensive cost of query time and storage space, and thus are not applicable in real-time applications, e.g., route recommendations in a self-driving car. This motivates us to develop a new Adaptive Monte Carlo Tree Search algorithm, namely A-MCTS. A-MCTS dynamically adjusts the priority of historical records that are used in Monte Carlo tree search to improve the performance and reduce the size of required “discovery experience”. In addition, A-MCTS proposes a new adaptive replay memory structure that can store important historical records based on graph properties and optimize the consumption of storage space. The experimental results on ten real-world dynamic graphs demonstrate that the proposed A-MCTS outperforms the state-of-the-art methods in terms of both efficiency and effectiveness.