fixed gaussians + reward function

[gandalfsaxe]

1. Gaussians missing parenthesis around sigma**2. Tuned Gaussian parameters slightly so it still looks like old example.
2. Switched around indices in G in plot title to get correct reward function, i.e. G[int(w[1]), int(w[0])] instead of G[int(w[0]), int(w[1])].

[cfong32]

I don't quite understand one point:
Why should we use G[int(wi[1]), int(wi[0])] (first [1] then [0]) in calculating the rewards? As far as I can see, wi[0] are corresponding to the x-coordinates and wi[1] are to the y-coordinates. Shouldn't we have the reward function written as G(x, y)? Why was it put as G(y, x)?

[gandalfsaxe]

I can't remember it clearly anymore, but you can see that further up he defines:

R = np.array([G[int(wi[1]), int(wi[0])] for wi in wp])

So that's of course also the way it should be when displayed in the plot.

I think the x/y mix up stems from the np.meshgrid that has 0,0 in upper left corner and positive x is right, positive y is down.