✅ UNIT – 1
Choose the Correct Alternative
1. What is deep learning?
a) A type of machine learning algorithm
b) A branch of artificial intelligence
c) A technique for training neural networks
d) All of the above
2. Which of the following is not a popular deep learning framework?
a) TensorFlow
b) PyTorch
c) Keras
d) Scikit-learn
3. What is the purpose of an activation function in a neural network?
a) It introduces non-linearity to the network
b) It determines the output of a neuron
c) It helps in backpropagation
d) All of the above
4. What is the role of gradient descent in deep learning?
a) To minimize the loss function
b) To find the optimal weights for the neural network
c) To update the network's parameters
d) All of the above
5. What is the vanishing gradient problem?
a) When the gradients in a deep neural network become extremely small
b) When the gradients in a deep neural network become extremely large
c) When the gradients in a shallow neural network become extremely small
d) When the gradients in a shallow neural network become extremely large
6. What is backpropagation used for in deep learning?
a) To calculate the gradients for updating the network's parameters
b) To propagate errors from the output layer to the input layer
c) To train a neural network
d) All of the above
7. Which of the following is a common loss function used for binary classification in
deep learning?
a) Mean Absolute Error (MAE)
b) Mean Squared Error (MSE)
c) Binary Cross-Entropy
d) Categorical Cross-Entropy
8. What is the purpose of dropout regularization in deep learning?
a) To reduce overfitting
b) To increase the model's capacity
c) To improve the training speed
d) To handle imbalanced datasets
9. What is an epoch in deep learning?
a) The number of layers in a neural network
b) The number of training examples in a dataset
c) The number of times the entire dataset is passed through the neural network
during training
d) The number of neurons in a layer
10. Which activation function is commonly used in the hidden layers of a deep neural
network?
a) ReLU (Rectified Linear Unit)
b) Sigmoid
c) Tanh
d) Softmax
Fill in the Blanks
1. Learning useful representations from data is the primary goal of Deep Learning.
2. Perceptron learning algorithm is the sequence of the given tasks in a perceptron.
3. The derivative of a sigmoid function is σ(x)(1 − σ(x)).
4. Early stopping is a technique used to prevent overfitting.
5. A perceptron takes a vector of real-valued inputs, calculates a linear combination of
these inputs, then outputs a binary value.
6. The purpose of data normalization in deep learning is to scale the data to a common
range.
7. Nesterov’s Accelerated Gradient Descent (NAG) differs from traditional momentum
by computing the gradient at the look-ahead position rather than the current one.
8. L2 regularization is used to regularize the network and prevent overfitting.
9. Dropout is a regularization technique that randomly sets a fraction of neurons to
zero during training.
10.Variants like Leaky ReLU and Parametric ReLU are used to address the issue of
dead neurons in ReLU.
✅ UNIT – 2
Choose the Correct Alternative
1. What is the purpose of a convolutional layer in a convolutional neural network
(CNN)?
a) To reduce the dimensionality of the input
b) To extract spatial features from the input
c) To classify the input data
d) To apply non-linear transformations to the input
2. What is the purpose of a recurrent neural network (RNN)?
a) To handle sequential data
b) To classify images
c) To perform dimensionality reduction
d) To generate synthetic data
3. Which type of RNN architecture is used to address the vanishing gradient problem?
a) Long Short-Term Memory (LSTM)
b) Gated Recurrent Unit (GRU)
c) Simple RNN
d) Bidirectional RNN
4. Which deep learning technique is used for generating new, realistic data samples?
a) Generative Adversarial Networks (GANs)
b) Convolutional Neural Networks (CNNs)
c) Reinforcement Learning
d) Transfer Learning
5. Which deep learning technique is used for sequence-to-sequence tasks, such as
machine translation?
a) Attention Mechanism
b) Convolutional Neural Networks (CNNs)
c) Transfer Learning
d) Autoencoders
6. Which deep learning technique is used for unsupervised feature learning?
a) Autoencoders
b) Convolutional Neural Networks (CNNs)
c) Recurrent Neural Networks (RNNs)
d) Reinforcement Learning
7. Which deep learning technique is used for learning latent representations from
unlabelled data?
a) Self-supervised learning
b) Reinforcement Learning
c) Transfer Learning
d) Unsupervised Learning
8. The main role of softmax activation function in a neural network's output layer is
a) To introduce non-linearity
b) To normalize the output into probability distribution
c) To prevent vanishing gradients
d) To reduce dimensionality of data
9. Which optimization algorithm is commonly used to train deep learning models?
a) Gradient Descent
b) K-Means
c) Genetic Algorithm
d) Adam
10. Which deep learning model architecture can be used for anomaly detection and
data reconstruction?
a) GAN
b) LSTM
c) Autoencoder
d) VAE
Fill in the Blanks
1. Kernel or filter is also known as a feature detector.
2. The final output from the series of dot products from the input and the filter is known
as a feature map.
3. Image classification is the major application of Convolutional Neural Networks.
4. The main function of pooling layer in CNN is to reduce the spatial dimensions of
feature maps.
5. Activation layer is commonly used in neural networks to add non-linearity.
6. Recurrent Neural Network (RNN) architecture has feedback connections.
7. The memory cell in LSTM has two states: cell state and hidden state.
8. GRU is Gated Recurrent Unit.
9. In VAE’s reconstruction error measures the difference between the original input
data and the reconstructed output.
10.GANs consist of a Generator and a Discriminator.
✅ UNIT – 3
Choose the Correct Alternative
1. Image segmentation is the process of dividing an image into multiple
a) Objects
b) Pixels
c) Regions
d) Frames
2. ________________ Segmentation assigns each pixel a class label, making it useful
for tasks like medical imaging.
a) Instance
b) Semantic
c) Object
d) Edge-based
3. The ________________ network is a popular deep learning model used for image
segmentation tasks.
a) Faster R-CNN
b) YOLO
c) U-Net
d) VGG-16
4. In image segmentation, ________________ methods group pixels based on
intensity values or colors.
a) Region-based
b) Object-based
c) CNN-based
d) Classification-based
5. ________________ is a deep learning model that detects multiple objects in
real-time.
a) U-Net
b) Mask R-CNN
c) YOLO
d) Autoencoder
Fill in the Blanks
1. YOLO is a one-stage object detection model that divides the image into a grid and
predicts bounding boxes directly.
2. The CNN–RNN model is commonly used for automatic image captioning.
3. Attention mechanism is a technique that helps image captioning models focus on
different parts of an image while generating text.
4. The BLEU score is commonly used to evaluate machine translation quality in
automatic image captioning.
5. The MS COCO dataset is widely used for training automatic image captioning
models.