Machine Learning Question:

conceptually we know two cases being far away from each other or close to each other with regard to the values of their features. A standard method of measuring such distances when the features are numerical in value is through the Euclidean distance, which is explained through the following numerical example: (There is an intermediate step of standardization of the features, the details and importance of which will discuss later. We will ignore that step for now) Suppose we have two people whose age, weight and height are:

case	age	height	weight
1	35	67	180
2	28	70	171

The Euclidean distance between the two cases based on these 3 features is defined as

square root (35 28)2 + (67 70)2 + (180 171)2 = 139

Let us apply this concept to the problem of identity verification. Suppose that a companys facial recognition algorithm converts a face picture into a set of 4 features ( current algorithms actually convert a picture into about 2000+ or 4000+ features). Suppose, further, that the company has a database of ID photos of Jill, Mary and Simon with associated feature values given as follows:

person	Feature 1	Feature 2	Feature 3	Feature 4
Jill	0.9	2.2.	0.5	-1.2
Mary	1.3	2.7	0.2	-1.8
Simon	0.7	1.8	0.9	-0.9

A new photo is presented to the algorithm, and it converts it to its feature values as

	Feature 1	Feature 2	Feature 3	Feature 4
New Photo	1	2.6	0.1	-0.7

Compute the Euclidean distance of the new photos feature values from those of the ID photos of Jill, Mary and Simon.

The algorithm considers a new photo as a match with one in its database if the Euclidean distance between the two is less than 0.5. Based on this rule, would the algorithm verify the new photo as belonging to any of the three people in its database?