Question about your logic to pick target

[Dongyang7]

I just wonder why the logic for picking a target on the screen is like picking a point yourself(by the network) and then a rectangular which has random height and random width and locates close to the point will be the 'target', is this the best way to model this? Did you adapted it from other's code?

[simonmeister]

@Dongyang7 In the original DeepMind paper, functions which require rectangle selection used two different spatial output layers for the top left and bottom right position (if that was what you meant). You can take a look at https://github.com/simonmeister/pysc2-rl-agents/tree/master/rl for a example implementation (WIP) that tries to be close to the deepmind paper.