1) Write a program to demonstrate the working of different activation functions like
sigmoid, tanh, ReLU & softmax to train a neural network.
Ans:
2)a) Design a single unit perceptron for classification of linearly separable binary dataset
without using pre-defined models. Use the perceptron from sklearn
Ans:
2)b) Identify the problems in single unit perceptron using AND, OR, XOR data and analyze
the results.
Ans:
3) Build a deep feed forward ANN by implementing the back propagation algorithm and
test the same using appropriate datasets. Use the no. of hidden layers greater than or
equal to 4.
Ans:
Output:
Epoch 1/10 15/15 ━━━━━━━━━━━━━━━━━━━━ 1s 1ms/step - accuracy:
0.2978 - loss: 1.1053
Epoch 2/10 15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 1ms/step - accuracy:
0.4320 - loss: 1.0677
Epoch 3/10 15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 2ms/step - accuracy:
0.5912 - loss: 0.9875
Epoch 4/10 15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 1ms/step - accuracy:
0.6311 - loss: 0.8460
Epoch 5/10 15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 1ms/step - accuracy:
0.5881 - loss: 0.7046
Epoch 6/10 15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 1ms/step - accuracy:
0.7379 - loss: 0.4710
Epoch 7/10 15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 1ms/step - accuracy:
0.6544 - loss: 0.4664
Epoch 8/10 15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 1ms/step - accuracy:
0.6476 - loss: 0.4563
Epoch 9/10 15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 1ms/step - accuracy:
0.6413 - loss: 0.4413
Epoch 10/10 15/15 ━━━━━━━━━━━━━━━━━━━━ 0s 2ms/step - accuracy:
0.6204 - loss: 0.4252
Test Accuracy: 70.00%
4) Design and implement a CNN model with 4+ layers of convolution to classify
multicategory image datasets. Use the concept of regularization and dropout while
designing CNN model. Use the fashion MNIST dataset. Record the training accuracy
and test accuracy corresponding to the following architecture:
a) Base model
b) Model with L1 regularization
c) Model with L2 regularization
d) Model with dropout
Ans:
5) Design and implement an image classification model to classify a dataset of images using
deep feed forward neural network. Record the accuracy corresponding to the no. of
epochs. Use MNIST dataset.
Ans:
Output:
6) Implement Bi-directional LSTM for sentiment analysis on movie reviews.
Ans:
7) Implement the standard VGG16 and 19 CNN architecture model to classify
multicategory image dataset and check the accuracy.
Ans:
