In order to predict target words given a context word, we need to perform forward propagation.
First, we multiply the input with the input to hidden layer weight :
We know each of the input words is one-hot encoded, so when we multiply with , we basically obtain the row of to . So, we can directly write as follows:
basically ...