Machine-Learning-Collection/ML/algorithms/svm/__pycache__/knn.cpython-37.pyc

B
<00>ˢ^<5E><00>@sldd
lZ
Gdd<03>d<03>Zedkrhe
jjdd<06>

e
jddd	<09>Ze
<01>d
<EFBFBD>
Zedd<0C>
Z	e	<09>
ee<08>
e	jedd
<0A>Zd
S)<0E>Nc@sFeZ
dZd
d<02>Zdd<04>Zddd<07>
Zdd	<09>Zd
d<0B>Zdd
<0A>Zdd<0F>Z	dS)<12>KNearestNeighborcCs|
|_d
|_
dS)Ng:<3A>0<EFBFBD><30>yE>)<02>
k<>eps)<02>selfr<00>r<00>+C:\Users\aladd\Desktop\ML_videos\svm\knn.py<70>__init__s

zKNearestNeighbor.__init__cCs|
|_||_
dS)
N)<02>X_train<69>y_train)r<00>
X<>
yrrr<00>trains

zKNearestNeighbor.trainr
cCs<|d
kr|<00>|
<01>
}n|dkr(|<00>
|
<01>
}n
|<00>|
<01>
}|<00>|<03>
S)Nr
<00>
)<03>compute_distance_vectorized<65>compute_distance_one_loop<6F>predict_labels)r<00>X_test<73>	num_loops<70>	distancesrrr<00>predicts



zKNearestNeighbor.predictcCs<>|
jd
}|j
jd
}t<02>||f<02>
}xbt|<02>
D]V}xPt|<03>
D]D}t<02>|jt<02>|
|dd<02>f|j
|dd<02>fd<00>
<00>
|||f<q<Wq.W|S)zp
        Inefficient naive implementation, use only
        as a way of understanding what kNN is doing
        r
N<>)<08>shaper	<00>np<6E>zeros<6F>range<67>sqrtr<00>sum)rr<00>num_test<73>	num_trainr<00>
i<>
jrrr<00>compute_distance_two_loopss



Hz+KNearestNeighbor.compute_distance_two_loopsc
Csv|
jd
}|j
jd
}t<02>||f<02>
}xLt|<02>
D]@}t<02>|jtj|j
|
|dd<02>fddd<05><00>
||dd<02>f<q.W|S)z<>
        Much better than two-loops but not as fast as fully vectorized version.
        Utilize Numpy broadcasting in X_train - X_test[i,:]
        r
Nrr)
<01>axis)rr	rrrrrr)rrrrrrrrrr)s


@z*KNearestNeighbor.compute_distance_one_loopcCsXtj
|
d
ddd<04>}tj
|jd
ddd<04>}t<00>|
|jj<04>}t<00>|j|d
||j<00>
S)a<>

        Can be tricky to understand, we utilize heavy
        vecotorization as well as numpy broadcasting.
        Idea: if we have two vectors a, b (two examples)
        and for vectors we can compute (a-b)^2 = a^2 - 2a (dot) b + b^2
        expanding on this and doing so for every vector lends to the 
        heavy vectorized formula for all examples at the same time.
        rrT)r"<00>keepdims)rrr	<00>dot<6F>
Trr)rrZX_test_squaredZX_train_squaredZtwo_X_test_X_trainrrrr8s	

z,KNearestNeighbor.compute_distance_vectorizedcCsn|
jd
}t
<01>|<02>
}xTt|<02>
D]H}t
<01>|
|dd<00>f<00>
}|j|d|j<06><00>t<08>
}t
<01>	t
<01>
|<06>
<01>
||<qW|S)Nr
)rrrr<00>argsortr
r<00>astype<70>int<6E>argmax<61>bincount)rrr<00>y_predr<00>	y_indicesZk_closest_classesrrrrHs






zKNearestNeighbor.predict_labelsN)
r
)
<EFBFBD>__name__<5F>
__module__<EFBFBD>__qualname__rr
rr!rrrrrrrrs

r<00>__main__r)
<01>seedzexample_data/data.txt<78>
,)
<01>	delimiterzexample_data/targets.txt<78>)
r)
r)
<0A>numpyrrr-<00>randomr1<00>loadtxtrr<00>KNNr
rr+rrrr<00><module>
sP