Researchers have developed a number of strategies for solving a stochastic puzzle like the Minesweeper for artificial intelligence (AI) agent.
For decades, game-solving efforts have focused exclusively on solving two-player games (i.e., board games such as checkers, games like chess, etc.), where some artificial intelligence (AI) is used to correct game results. And can be effectively predicted. Collect a large number of search techniques and gameplay statistics.
However, such methods and techniques cannot be applied directly to the domain that solves the puzzle because puzzles are usually played alone (single-player) and have unique features (such as a stock stick or hidden information). ). So, the question arises of how the AI technique can maintain its performance to solve a two-player game but instead apply to the single-agent puzzle.
For years, puzzles and games were considered interchangeable or part of one another. Honestly, it can’t happen all the time. From a real-world perspective, the game is something we have to deal with every day. Dealing with the unknown. For example, making the right decision is unknown (ie getting married) or making the wrong decision (ie quitting the job) or not doing it at all (ie regretting what will happen).
Meanwhile, the ‘puzzle’ is something that was known, and there is even something hidden that remains to be uncovered. For example, one such well-known case would be the discovery of a ‘Wonder’ material like graphene and its many capabilities that are yet to be used commercially and widely. Then again, how and what is the boundary between ‘puzzle’ and ‘game’ in the context of solving puzzles?
In Japan, at the Advanced Institute of Science and Technology (JAIST), in Japan, Professor Hiroyuki Ida, and colleagues attempted to answer these two questions in their latest study, published in the journal Knowledge-Based Systems. The research study focuses on two key contributions:
(1) the resolution of a puzzle in a single-agent game context through the Minesweeper testbed and
(2) using a unified structure of four strategies called PAFG. Suggest a new artificial intelligence (AI) agent. Taking advantage of the known and unknown information of the solver Minesweeper puzzle, the proposed solver had a better performance in solving this puzzle than the latest studies.
The statistics show AI strategies that use knowledge-based strategies to deal with anonymous information while adopting data-driven strategies to use the known information of the Minesweeper puzzle. Outcomes establish a solvable condition in a single-player stockists puzzle that is canonical for a wide range of real-world problems. Credit: HeroWiki Idea from JAIST
The researchers adopted an AI agent consisting of two knowledge-driven strategies and two data-driven strategies to make the best use of the known and unknown information of the present decision so that the subsequent decision could be made. The best guess is. As a result, a single-agent stochastic puzzle, like the Minesweeper, can draw a line between puzzle-solving and gameplay.
Such a situation plays a particularly important role in real-world problems where the boundary between known and unknown is usually blurred and difficult to identify. As Professor Ida remarks: “With the AI ?? agent’s ability to increase puzzle-solving efficiency, the range of solutions becomes clearer.
Allowing compliments, which are commonly found in many real-life situations, such as determining a high-stakes investment, estimating the risk level of an important decision, etc. In short, we all have our own mine. Living in the world of sweepers, trying to figure out the way forward by avoiding the ‘bombs’ in their lives.
There was a lot of uncertainty with the rapid development of existing technology and new models of computing available (ie IoT, cloud-based services, edge computing, neuromorphic computing, etc.).
This situation can be true for people (ie technical capability), community (ie acceptance of technology), society (ie culture and routine), and even the national level (ie change in policy and rules). “Everyday human activity involves a lot of ‘game’ and ‘puzzle’ situations. The risk can be minimized and the benefit of the known can be maximized, “said Ms.
Chang Liu, the study’s lead author. “Such an achievement is achieved by eliminating knowledge-driven techniques, AI technology, and measurable uncertainties (such as winning rate, success rate, growth rate, etc.) while making this puzzle interesting Despite the challenges. ”
Reference: “Single Agent Stochastic Puzzle Solving: A Case Study with Minesweeper” by Chang Liu, Shunki Huang, Gao Naing, Muhammad Noor Akmal Khalid, and Herowiki Ida, March 28, 2022, Knowledge-Based Systems.
DOI: 10.1016 / j.knosys.2022.108630