Improved Feature Learning: A Maximum-Average-Out Deep Neural Network for the Game Go

Figure 5

Go boards during games between programs using the ResNet18 and MAO algorithms. In this figure, the numbers on the stones indicate the steps that each player plays and red boxes highlight the stones, represented by a number without a stone present, taken by the opponent in the enclosed area. (a) MAO (B) vs. Resnet18 (W), (b) Resnet18 (B) vs. MAO (W), and (c) Resnet18 (B) vs. MAO (W).

(a)

(b)

(c)