Constants in the intensity function

[YuanYuan98]

The paper states that , where is a neural network with a softplus nonlinearity applied to the output, to ensure the intensity is positive. However, the implementation is not consistent, where a sigmoid activation function is used. In addition, it is hard to understand the constants 2.0 and 50 in the following intensity calculation function.

def get_intensity(self, tpp_state): 
    return torch.sigmoid(self.intensity_fn(tpp_state[..., :self.hdim]) - 2.0) * 50