From T2 FS-Based MoGoTW System to DyNaDF for Human and Machine Co-learning on Go

This chapter describes the research from T2 FS-based MoGoTW system to DyNamic DarkForest (DyNaDF) open platform for human and machine co-learning on Go. A human Go player’s performance could be influenced by some factors, such as the on-the-spot environment as well as physical and mental situations of the day. In the first part, we used a sample of games played against machine to estimate the human’s strength (Lee et al. in IEEE Trans Fuzzy Syst 23(2):400–420, 2015 [1]). The Type-2 Fuzzy Sets (T2 FSs) with parameters optimized by a genetic algorithm for estimating the rank was presented (Lee et al. in IEEE Trans Fuzzy Syst 23(2):400–420, 2015 [1]). The T2 FS-based adaptive linguistic assessment system inferred the human performance and presented the results using the linguistic description (Lee et al. in IEEE Trans Fuzzy Syst 23(2):400–420, 2015 [1]). In March 2016, Google DeepMind challenge match between AlphaGo and Lee Sedol in Korea was a historic achievement for computer Go development. In Jan. 2017, an advanced version of AlphaGo, Master, won 60 games against some top professional Go players. In May 2017, AlphaGo defeated Ke Jie, the top professional Go player, at the Future of Go Summit in China. In second part, we showed the development of computational intelligence (CI) and its relative strength in comparison with human intelligence for the game of Go (Lee et al. in IEEE Comput Intell Mag 11(3):67–72, 2016 [2]). Additionally, we also presented a robotic prediction agent to infer the winning possibility based on the information generated by DarkForest Go engine and to compute the winning possibility based on the partial game situation inferred by FML assessment engine (Lee et al. in FML-based prediction agent and its application to game of Go, 2017 [3]). Moreover, we chose seven games from 60 games to evaluate the performance (Lee et al. in FML-based prediction agent and its application to game of Go, 2017 [3]). In this chapter, we extract the human domain knowledge from Master’s 60 games for giving the desired output. Then, we combine Particle Swarm Optimization (PSO) and FML to learn the knowledge base and further infer the game results of Google AlphaGo in May 2017. The experimental results show that the proposed approach is feasible for the application to human and machine co-learning on Go. In the future, powerful computer Go programs such as AlphaGo are expected to be instrumental in promoting Go education and AI real-world applications.