It looks like the 5th line in Summing up the training scheme# has mixed up the target variable ‘t’ with iteration index (which should also be e.g. j and j - 1.
Please check and confirm.
The calculated weight is absolutely fine where number of iterations are shown using i variable.