
上QQ阅读APP看书,第一时间看更新
Notation
Let the data be of the form , where:
,
(number of classes = 2 because it's a binary classification)
is 'n' dimensional, that is,
(refers to the preceding diagram)
The number of training examples is m. Thus the training set looks as follows:
.
m = size of training dataset.
And, since
, where, each
.
Therefore,
is a matrix of size n * m, that is, number of features * number of training examples.
, a vector of m outputs, where, each
.
Parameters : Weights
, and bias
,
where
and
is a scalar value.