Tensorflow

>
-
Pleas ??

an-dimensional
Tensor
- NN
11 - Matimliplication
Matric
Gradient
NN
=>
calculation

in-00
Numerical
NumPy
a
e

- Python

Dob
model ent
 I

sportat
formation)
m neens

↑
layeTransformat
>
-
-
a -

Mediction
-
?
S

Fee
Optimizer
r
X L
function

-
was

Th
 Price Prediction

⑧
house o -1
- ② -
1-10
bedroom I o -1
① No of - -
5500
=

so
Size
② House
-

Mates +b
NN - x3

wat
--

⑧ z =
w
⑧
.
Xi +

Pat
Sir
 function Features
Activation
-
-

ent
different
mil
AF(2) e
can

weight
Tan ,
weight eleg
Signoid ,

-
Q

#
I
- Re

=
- 1(500)
Ho
000)
- 1500 ,
0-255 =
of
-
255
Range
~
Data Type

T
I normalization
②

D ·
E
Outlier
Data point
from
that
is

rest

B
I10 ,
20 ,

Ab o -1
v. different data
of your
↑ aug
. ..
0
0 .

mean +
standard
·

!
·
E = deviation
-
sir
variance
- H)
2

E(x
-

+
 Chris
Burd
X -
H
=
-
-
-

2-scoll
Ne
-
↑ Nas
standardization
Scaling -

10 ,
1) o
Range
=
Mean
1) Ext
>
- I I
(f , X = sea =

X =
Xi xni
>
outliers
be
handled
outliers
-

can

I Not good s
Range
-Data
e
Al
0-255
0-1000-1
Tempa

b outliers
>
- sensitive
 Pre-processing
=
running
D
lowercase e
Text
Tokenization
2
run
- &
legten
00 same

3
-00
- Padding
O
a
- words
step "

-
Remove
~
I am

the

to
fired Sine 0-
Scale 0-255

Image
Denoisig conversion
Grand os Grayscale
&
Dewaising
Audio umoval
-
&
Silence
Frequency
=
Amplitude
 Null
Tabular I Handle valus
-

Remove
=
-
Man
Transform
= log
I

Normalization irrelevant
T
woise , deplicates &
features
- Removing
Reduction
>
- Dimensiencin
 Educosys >
- Numbers ?
2
love
000

11
-10 ,
Educosys cat 000
Y
love - Late 100 ,

I dog
-

- a 5

① Tokenization -
O Ratiy
& 6
Like
Label 3
encoding
D
>
-

rected dur

I Binary plant,

/
hot
Encoding one
of
②
-
No
encoding
·

0
[1 , classe
ConE) 0 0] binary
90 no of
,
1
ucen
07
, .
,
=

0 .
1
, of , in
10 ,
0
, 7 laugen a
recher

of close
0
10 ,

③ Embedding Relationship
,

= no

content,
Learn words ver
between II
Sever Neuralwork
 > I 0 >

it
...

1. --
-- - 3
love
]
Educosys 1. ..
--

I 9
:

##
 Do
(0
3
2
cat .

nor
5)
.

0 .

-
0 . 5
1 0

sand (0
8
.

0
1
.

-0 30
7)
. .

0 9
9
.

20
.

. 8
my

H
 los

Accuracy
-

)
-

E
Percentage
correct predictions pdicted
% of Loss
Functoru
predicts
Di
+100 2

correct
= Nof
-

Total
Predicts
98
%
&
000
10 - ,

- 200
ar
wrong

ar
correct

More
perdictions
>
- values
Accuracy to
true
↑
Nearer
e
to
Loss
Cresification
-

Entropy
Entropy
·
Binau Dependenttargets
Binary & on

T
Price
House
Prediction

I
o
and
between
binaya
Regression
-

Indiction
Price
Stock
Rs
10-500 ad
MAE
e
5-1
Price
MAE He
error

Prediction k$
Price 000
never-00
,

kS Good
500
·use
MAE Bad .

up
2000k
en
voc
06
Temperature
- Maee3o2
e 10
 00

:
-

e
4 5
-3 , ,

3
4 5,

-5
,

6 .
7
,

7
8
- 6 ,
.

8 ,
9
~
1 .

3.
namel %)
·
 ,
6 2 8 9
5

3 Paddig
1 , , ,
, ,

t
Padding T
2
, '
D 9
- 9 size
1
2
, 3 same
-
4 olp -

23 .
T as ilp
acces
2
Padder e
4
,
3 .
-

5, b Add
y , T
7
6,
5 , -
8
7
6 ,
,
T

7 8, 9
, T
10
8, 9
,
11210

· 456789
T
-
i z T
3
,

2 ,
1 , T

i r

-
3
5
y ,
.

.
5
6 , 7
6 ,
2

81
.

89 -

1 -
,

5.
9
T
9