[josecyc]

Haven't read the paper, but it's still unclear how the mechanism of one shot learning works. If the weights are not being updated, how is it "learning"?