THAKUR COLLEGE OF

SCIENCE AND COMMERCE

Affiliated to University of Mumbai 2022 – 2023

PRACTICAL JOURNAL OF
DEEP LEARNING

SUBJECT GUIDE
ANAND UPADHYAY

SUBMITTED BY
SHONA KANOJIYA
ROLL NO: 178

SUBMITTED IN PARTIAL FULFILLMENT OF THE

REQUIREMENTS

FOR QUALIFYING M. Sc.IT PART II

SEMESTER IV EXAMINATION)

1
 CERTIFICATE OF APPROVAL

This is to certify that Ms. Shona Kanojiya student of “Master of

Science (Information Technology)” of “Thakur College of Science and

Commerce”.

Roll No. 178 has successfully completed and submitted the

Practical & Assignment entitled “Deep Learning” in the partial fulfilment

as per the syllabus defined by the University of Mumbai in the academic

year 2022- 2023.

It is further certified that the student has completed all the required

phases of the practical & assignment.

_______________________
HEAD OF DEPARTMENT

_______________________ _______________________
PROFESSOR INCHARGE EXTERNAL EXAMINER

2
 INDEX

SR. PRACTICAL NAME DATE PAGE SIGN

NO NO
1 Write a program to read and pre-process the data. 22/01/22 04

2 Write a program to demonstrate various operation 19/03/22 08

related to tensor, matrix and vector.
3 Write a program to implement multi perceptron 28/03/22 12
using TensorFlow.

4 Write a program to implement concept of 28/03/22 14

convolution operation.

5 Write a program to implement convolution with 28/03/22 17

various layers.

6 Write a program to display and plot various 06/02/23 20

performance related parameter for CNN.

7 Write a program to implement RNN. 14/02/23 22

8 Write a program to show role of different 15/01/23 23

optimization algorithm for underfitting and
overfitting.

9 Write a program to implement deep neural 01/02/23 26

network and save trained model for future use.

10 Write a program to implement deep neural 28/01/23 29

network with various pretrained model.

3
 Practical 01
Aim: Write a program to read and pre-process the data.

1. Read the data.

import pandas as pd
df=pd.read_csv("/content/Train.csv")
df

2. Disable header and footer of data.

csv_file=df.to_csv(header=None)
print(csv_file)

4
3. Plot bar graph, pie chart, line diagram and scatter plot.

plt.bar(df["Width"],df['Height'])
plt.show()

plt.scatter(df["Width"],df['Height'])
plt.show()

plt.plot(df["Width"],df['Height'])
plt.show()

5
plt.pie(df['Height'])
plt.show()

import numpy as np
plt.bar(df["Width"], 0.4, label = 'Width')
plt.bar(df["Height"], 0.4, label = 'Height')
plt.legend()
plt.show()

4. Apply filter on the data such as >,<,>=,between ,in and range.

rslt_df = df[df['Width'] > 70]

rslt_df.head(10)

6
low = [x * 0.01 for x in range(0, 10)]

low

x["latitude"].between(2, 4, inclusive = True)

7
 Practical 02
Aim: Write a program to demonstrate various operation related to tensor, matrix and vector.
Source Code & Output:
1. Tensor operation.
import tensorflow as tf
import numpy as np
a=tf.constant([[1,2,3],[4,5,6],[7,8,9]])
a

b=tf.constant([[3,2,3],[4,5,6],[1,8,2]])
b

c=tf.add(a,b)
c

print(tf.matmul(a,b))

print(tf.reduce_max(a))

print(tf.reduce_max(b))

print(tf.math.argmax(b))

a=tf.ones([3,3])
a
print(tf.nn.softmax(a))

8
 a=tf.zeros([3,3])
a

A = tf.constant([[1, 2], [3, 4]]);

B = tf.reshape(A,[4, 1]);
B

import tensorflow as tf
a=tf.range(3, 19, 3)
a

tensor = tf.constant([[1, 2, 3], [4, 5, 6]])

tf.zeros_like(tensor)

tf.random.shuffle(tensor, seed=2, name=None)

2. Matrix operation.
import numpy as np
mx1 = np.array([[5, 10], [15, 20]])
mx2 = np.array([[25, 30], [35, 40]])

9
 print("Matrix1 =\n",mx1)
print("\nMatrix2 =\n",mx2)
print ("\nAddition of two matrices: ")
print (np.add(mx1,mx2))
print ("\nSubtraction of two matrices: ")
print (np.subtract(mx1,mx2))
print ("\nMatrix Division: ")
print (np.divide(mx1,mx2))
print ("\nMultiplication of two matrices: ")
print (np.multiply(mx1,mx2))
print ("\nThe summation of elements=")
print (np.sum(mx1))
print ("\nThe Transpose =")
print (mx1.T)
print("\ninner:", np.inner(mx1, mx2))
print("dot:", np.dot(mx1, mx2))
det = np.linalg.det(mx1)
print("\nDeterminant:", np.round(det))
rank = np.linalg.matrix_rank(mx1)
print("\nRank:", rank)
print("\nTrace:", mx1.trace())
print("Trace:", sum(mx1.diagonal()))

3. Vector operation.
import numpy as np
list1 = [10,20,30,40,50]
list2 = [11,12,13,14,15]

10
vtr1 = np.array(list1)
vtr2= np.array(list2)
print("vector1:")
print(vtr1)
print("vector 2:")
print(vtr2)
print("Addition of two vectors: ",vtr1+vtr2)
print("Subtraction of two vectors: ",vtr1-vtr2)
print("Multiplication of two vectors: ",vtr1*vtr2)
print("Division of two vectors: ",vtr1/vtr2)
print("Dot product of two vectors: ",vtr1.dot(vtr2) )

11
 Practical 03
Aim: Write a program to implement multi perceptron using TensorFlow.
Source Code & Output:
import tensorflow as tf
import numpy as np
from tensorflow import keras
import matplotlib.pyplot as plt
(x_train, y_train), (x_test, y_test) = tf.keras.datasets.mnist.load_data()
x_train = x_train.astype('float32')
x_test = x_test.astype('float32')
gray_scale = 255
x_train /= gray_scale
x_test /= gray_scale
print("Feature matrix:", x_train.shape)
print("Target matrix:", x_test.shape)

fig, ax = plt.subplots(10, 10)

k=0
for i in range(10):
for j in range(10):
ax[i][j].imshow(x_train[k].reshape(28, 28),aspect='auto')
k += 1
plt.show()

12
model=keras.models.Sequential()
model.add(keras.layers.Flatten(input_shape=[28,28]))
model.add(keras.layers.Dense(256,activation="sigmoid"))
model.add(keras.layers.Dense(128,activation="sigmoid"))
model.add(keras.layers.Dense(10,activation="sigmoid"))
model.compile(optimizer='adam',loss='sparse_categorical_crossentropy',metrics=['accuracy'])
model.fit(x_train, y_train, epochs=10, batch_size=2000, validation_split=0.2)

results = model.evaluate(x_test, y_test, verbose = 0)

print('test loss, test acc:', results)

13
 Practical 04
Aim: Write a program to implement concept of convolution operation.
Source Code:
from keras.layers import Convolution2D, MaxPooling2D, Activation
from keras.models import Sequential
from numpy import asarray
import matplotlib.pyplot as plt
import cv2
img = cv2.imread('/content/download.jpg',cv2.IMREAD_GRAYSCALE)
plt.imshow(img, cmap = 'gray')
plt.show()
img.shape
img_batch = img.reshape(1,img.shape[0], img.shape[1],1)
img_batch.shape
img_batch.shape[1:]
model1 = Sequential()
model1.add(Convolution2D(1,(15,15) , padding = 'valid', input_shape=img_batch.shape[1:]))
#random weight initilaization
model1.summary()
conv_img = model1.predict(img_batch)
conv_img.shape
conv_img_show = conv_img.reshape(conv_img.shape[1], conv_img.shape[2])
print(conv_img_show)
plt.imshow(conv_img_show, cmap = 'gray')
plt.show()
model2 = Sequential()
model2.add(Convolution2D(1, (15,15) , padding = 'valid', input_shape=img_batch.shape[1:])

14
) # random weight initilaization
model2.add(Activation('relu'))
conv_img = model2.predict(img_batch)
conv_img_show = conv_img.reshape(conv_img.shape[1], conv_img.shape[2])
plt.imshow(conv_img_show, cmap = 'gray')
plt.show()
model4 = Sequential()
model4.add(Convolution2D(1, (15,15) , padding = 'valid', input_shape=img_batch.shape[1:])
) # random weight initilaization model4.add(Activation('relu'))
model4.add(MaxPooling2D(pool_size=(2,2)))
model4.summary()
conv_img = model4.predict(img_batch)
conv_img_show = conv_img.reshape(conv_img.shape[1], conv_img.shape[2])
plt.imshow(conv_img_show, cmap = 'gray')
plt.show()
Output:

15
16
 Practical 05
Aim: Write a program to implement convolution with various layers.
Source Code & Output:
import tensorflow as tf
from tensorflow import keras
tf.__version__
fashion_mnist=keras.datasets.fashion_mnist
print(fashion_mnist)
(x_train_full,y_train_full),(x_test,y_test)=fashion_mnist.load_data();
#x_train_full.shape
x_test.shape
x_validate,x_train=x_train_full[:5000]/255.0,x_train_full[5000:]/255.0
y_validate,y_train=y_train_full[:5000],y_train_full[5000:]
x_validate.shape

class_name=["T-
shirt/Top","Trouser","pullover","Dress","coat","sandal","shirt","sneaker","bag","Ankleboot"]

model=keras.models.Sequential([
keras.layers.Conv2D(64,7,activation="relu",padding="same",input_shape=[28,28,1]),
keras.layers.MaxPooling2D(2),
keras.layers.Conv2D(128,3,activation="relu",padding="same"),
keras.layers.Conv2D(128,3,activation="relu",padding="same"),
keras.layers.MaxPooling2D(2),
keras.layers.Conv2D(256,3,activation="relu",padding="same"),
keras.layers.Conv2D(256,3,activation="relu",padding="same"),
keras.layers.MaxPooling2D(2),

17
 keras.layers.Flatten(),
keras.layers.Dense(128,activation="relu"),
keras.layers.Dropout(0.5),
keras.layers.Dense(64,activation="relu"),
keras.layers.Dropout(0.5),
keras.layers.Dense(10,activation="softmax"),
])
model.summary()

model.layers[0]
hidden1=model.layers[1]

18
hidden1.get_weights()
model.compile(loss="sparse_categorical_crossentropy",optimizer="adam",metrics=["accuracy"])
history=model.fit(x_train,y_train,epochs=20,steps_per_epoch=5,validation_data=(x_validate,y_
validate))

19
 Practical 06
Aim: Write a program to display and plot various performance related parameter for CNN.
Source Code & Output:
import pandas as pd
import matplotlib.pyplot as plt
plt.figure(figsize=(8,5))
plt.plot(pd.DataFrame(history.history))
plt.grid(True)
plt.gca().set_ylim(0,1)
plt.show()

import pandas as pd
import matplotlib.pyplot as plt
pd.DataFrame(history.history).plot(figsize=(8,5))
plt.grid(True)
plt.gca().set_ylim(0,1)
plt.show()

20
21
 Practical 07
Aim: Write a program to implement RNN.
Source Code & Output:
import keras

from keras.datasets import mnist

from keras.models import Sequential

from keras.layers import Dense, Dropout, LSTM

from keras.optimizers import Adam

(X_train, y_train),(X_test, y_test) = mnist.load_data()

X_train = X_train.astype('float32') / 255.0

X_test = X_test.astype('float32') / 255.0

model=keras.Sequential()

model.add(LSTM(128, input_shape=(X_train.shape[1:]), return_sequences=True))

model.add(Dropout(0.2))

model.add(LSTM(128))

model.add(Dense(64, activation='relu'))

model.add(Dropout(0.2))

model.add(Dense(10, activation='softmax'))

model.compile( loss='sparse_categorical_crossentropy',optimizer='Adam',

metrics=['accuracy] )

history=model.fit(X_train,y_train, epochs=3,validation_data=(X_test, y_test))

model.evaluate(X_test, y_test, verbose=0)

22
 Practical 08
Aim: Write a program to show role of different optimization algorithm for underfitting and
overfitting.
Source Code & Output:
import numpy as np
import matplotlib.pyplot as plt
from sklearn.pipeline import Pipeline
from sklearn.preprocessing import PolynomialFeatures
from sklearn.linear_model import LinearRegression
from sklearn.model_selection import cross_val_score
def true_fun(X):
return np.cos(1.5 * np.pi * X)
np.random.seed(0)
n_samples = 30
degrees = [1, 4, 15]
X = np.sort(np.random.rand(n_samples))
y = true_fun(X) + np.random.randn(n_samples) * 0.1
plt.figure(figsize=(14, 5))
for i in range(len(degrees)):
ax = plt.subplot(1, len(degrees), i + 1)
plt.setp(ax, xticks=(), yticks=())
polynomial_features = PolynomialFeatures(degree=degrees[i], include_bias=False)
linear_regression = LinearRegression()
pipeline = Pipeline(
[
("polynomial_features", polynomial_features),
("linear_regression", linear_regression),

23
 ]
)
pipeline.fit(X[:, np.newaxis], y)
# Evaluate the models using crossvalidation
scores = cross_val_score(
pipeline, X[:, np.newaxis], y, scoring="neg_mean_squared_error", cv=10
)
X_test = np.linspace(0, 1, 100)
plt.plot(X_test, pipeline.predict(X_test[:, np.newaxis]), label="Model")
plt.plot(X_test, true_fun(X_test), label="True function")
plt.scatter(X, y, edgecolor="b", s=20, label="Samples")
plt.xlabel("x")
plt.ylabel("y")
plt.xlim((0, 1))
plt.ylim((-2, 2))
plt.legend(loc="best")
plt.title(
"Degree {}\nMSE = {:.2e}(+/- {:.2e})".format(
degrees[i], -scores.mean(), scores.std()
)
)
plt.show()

24
25
 Practical 09
Aim: Write a program to implement deep neural network and save trained model for future use.
Source Code & Output:
Save Model:
model=keras.models.Sequential([
keras.layers.Conv2D(64,7,activation="relu",padding="same",input_shape=[28,28,1]),
keras.layers.MaxPooling2D(2),
keras.layers.Conv2D(128,3,activation="relu",padding="same"),
keras.layers.Conv2D(128,3,activation="relu",padding="same"),
keras.layers.MaxPooling2D(2),
keras.layers.Conv2D(256,3,activation="relu",padding="same"),
keras.layers.Conv2D(256,3,activation="relu",padding="same"),
keras.layers.MaxPooling2D(2),
keras.layers.Flatten(),
keras.layers.Dense(128,activation="relu"),
keras.layers.Dropout(0.5),
keras.layers.Dense(64,activation="relu"),
keras.layers.Dropout(0.5),
keras.layers.Dense(10,activation="softmax"),
])
model.save('/content/drive/MyDrive/modelCNN.h')
from google.colab import drive
drive.mount('/content/drive')
Loading the data:

import tensorflow as tf

from tensorflow import keras

tf.__version__

26
mod=tf.keras.models.load_model('/content/drive/MyDrive/model1.h')

mod.summary()

fashion_mnist=keras.datasets.fashion_mnist

print(fashion_mnist)

(x_train_full,y_train_full),(x_test,y_test)=fashion_mnist.load_data();

27
res=mod.evaluate(x_test,y_test,batch_size=64)

28
 Practical 10
Aim: Write a program to implement deep neural network with various pretrained model.
Source Code & Output:
VGG Model:
import tensorflow as tf

from tensorflow import keras

from keras.applications.vgg16 import VGG16

from keras.utils import plot_model

fashion_mnist=keras.datasets.fashion_mnist

print(fashion_mnist)

(x_train_full,y_train_full),(x_test,y_test)=fashion_mnist.load_data();

from keras import models

base_model = VGG16(weights='imagenet')

model_vgg = models.Model(inputs=base_model.input, outputs=base_model.get_layer('flatten').o

utput)

model_vgg.summary()

29
model_vgg.compile(loss="sparse_categorical_crossentropy",

optimizer="sgd",

metrics=["accuracy"])

history=model_vgg.fit(x_train,y_train,epochs=20,validation_data=(x_validate,y_validate))

30
31