Fig. 3
From: Deep bidirectional intelligence: AlphaZero, deep IA-search, deep IA-infer, and TPC causal learning
Monte-Carlo tree search, CNneim-A search, Bi-Scout A*, and extensions. a A snapshot after circling four steps of MCTS for certain times. b A snapshot after running A* for a while. c A snapshot after running CNneim-A for a while. d A snapshot after running Bi-Scout A* for a while. e Deep learning network for estimating value and policy. f A spectrum of selection strategies by combining A* and MCTS. g Backward and forward revision of f values in OPEN. h The lost function is minimised for learning parameters \(\theta\)