1
DEEP LEARNING
by
Mrs.Balne Sridevi
M.Tech(Ph.D),MISTE,MCSI
Asst. Professor
Dept. of CSE
Balaji Institute of Technology and Science
Contents
2
 AI, ML and DL?
 Why Deep Learning
 Neural Network Models
 Schematic Diagram of Inception Model
 Activation Functions
 Pooling Techniques
 Demo
AI,ML & DL
3
4
Why Deep Learning?
5
6
Neural Network Models
7
 Artificial Neural Network (ANN)
 Convolution Neural Network (CNN)
 Recurrent Neural Network (RNN)
Neural Network Model
8
ANN Architecture
9
CNN Architecture
10
RCNN
11
Schematic Diagram of Inception Model
Activation Functions
13
 Sigmoid Activation Function
 ReLu Activation Function
14
15
Max Pooling
16
17
18
Average Pooling with stride 2
20
21
22
Basic Steps to Fit a Model
 Linear Regression
 NN Model
 Load Data
 Define a model
 Add Layers, Activation Function
 Compile Model
 Fit Model
23
Linear Regression
24
Scatter() And Show() Methods
 plt.scatter(x,y,color=‘r’)
Where r is red in color
 plt.show() is used to display the x and y
coordinates
25
How Neural Network Works?
26
How Neural Network Works?
27
Load Data
 Import pandas Library
 Create CSV File
 Load Data by using read_csv() method
 Syntax: data_frame = pd. read_csv(‘ file
name.csv’)
28
Define Model
 Import model from keras library
 Import layers from keras library
 Define model as model=Sequential()
29
Add Layers and Activation Functions
 model.add(Any_net_Inception_model(Number of
layers), activation Function=‘function_name’)
 Inception Models are Dense,Imagenet,Residual
Net V1, V2
30
Compile Model
 model.compile(loss,optimizer,metrics)
 Loss={binary_crossentropy (multi Label ) or
categorical_crossentropy (multi classes))
 Optimizers={adam or adagrad or sgd etc..}
 metrics={accuracy ,count etc..}
31
Fit a Model
 model.fit(x,y,epochs,batch_size)
32
33
34
35
36
37
38
39
40
41
42
43
44
A Deep Dream of Neural
Network
https://www.youtube.com/watch?v=sh-MQboWJug
45
Thank You
46

Deep learning ppt