Neural networks vs. tree search in puzzle solving

Summary form only given, as follows. Two puzzles are solved on neural networks, using the method of energy minimization, and a simple comparison is made for the actual speeds of neural networks and tree search algorithms. The authors treat the n-queens problem and the polyomino puzzle. These problems have been considered as good examples for tree search techniques in artificial intelligence, but they also have two convenient advantages as example problems for neural optimization. One advantage is that the minimum energy value can be known when the energy function is defined. The other advantage is that the efficiency of convergence can be estimated by noting only solutions corresponding to global minima because only solutions with the minimum energy value have meaning in the puzzles.<
>