The objective of this lab is to program and test two classication algorithms,

very simple but very eective: the K$-$Nearest Neighbor $($KNN$)$ algorithm and the

Classier Bayesian Naive $($CBN$).$ We are studying here only the simplest versions

of these algorithms. For this lab we will need to import sklearn and numpy. The

tests can be done on sklearn's predened data that comes with their class labels

$($target$),$ for example:

iris $=$ datasets.load$\_$iris $()$

X $=$ iris.data

Y $=$ iris.target

A Nearest neighbor

The Nearest Neighbor algorithm is a very simple classication algorithm which is

based on the following principle: the class of each test data $($to be classied$)$ must

be the class of the closest $($most similar$)$ data among the training data. List of

useful functions:

$-$ metrics.pairwise.euclidean$\_$distances: calculates distances between data.

$-$ argsort: returns the indices of the ordered vector

$-$ argmin, argmax: returns the indices of the minimum$/$maximum values

$-$ neighbors.KNeighborsClassifier: K Nearest Neighbors alg. of sklearn

$1.$ Create a TNN function $($X$,$ Y$)$ which takes X data and labels as input Y and

which returns a label, for each data, predicted from the nearest neighbor of

this data. Here we take each data, one by one, as data test and we consider

all others as learning data. That we allows to test the power of our algorithm

according to a validation method by cross validation of leave one out type.