Let's look at a detailed view of the logistic unit and the cost function. We are going to use the logistic unit in a different manner. Before feeding our data directly to the logistic unit, we will first calculate the difference between each of these units, and only then will we multiply this difference by the weight, and then fit it to the logistic unit. This can be expressed mathematically expressed as follows:

We take the difference of each of these units and then multiply it with some weights. Sum up all the values and apply it to the sigmoid function that could output zero and one.

We could complicate this formula by writing it in this form:

Here, we calculate the squared difference and then divide it by the sum. We then, sum up these values and apply it to a sigmoid function. 

We could keep on complicating this formula, but the ideology stays the same.