In the book by K.Du, "Neural Networks and Machine Learning", Springer, 2014, p.159", the Hopfied dynamics equation (in discrete network model) is given as
$$net_{i}(t+1) = \sum_{i=1}^{J}x_{i}(t)w_{ij} + \theta{i}$$
Where: $x_{i}(t + 1) = \theta(net_{i}(t + 1))$,
$net_{i}$ is the weighted net input of the i-th neuron, xi (t) is the output of the i-th neuron and $\theta$ the threshold of the neuron. The above Hopfield net consists of a total of $J$ perceptrons and t notates time-state in the network.
The net input of the perceptron (McCulloch-Pitts model) is given by
$$\sum_{i=1}^{n}x_{i}w_{i} - \theta$$
In this model, it seems perfectly logical to detract the threshold of the nuron from the total collected weight.
However why in the hopfield dynamics equation Du adds the threshold instead of substrating it?