Scribd
Upload
40 views

Deep Learning Exam Answers

Uploaded by

tanutv66
Copyright
© © All Rights Reserved
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
40 views

Deep Learning Exam Answers

Uploaded by

tanutv66
Copyright
© © All Rights Reserved
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 2

Deep Learning Examination - Answer Key

Part A: Multiple Choice Answers


1. b) To automate feature extraction and learning.

2. b) Neuron

3. b) Convolutional Neural Network (CNN)

4. c) To apply filters to extract local features from the image.

5. c) Reduces the spatial dimensions of the image by taking the maximum value in each
region.

6. b) Tanh

7. c) To update the weights using the gradient of the error.

8. a) Processes data sequentially and maintains memory of past inputs.

9. b) Recurrent Neural Network (RNN)

10. a) Prevents overfitting.

Part B: Short Answer Solutions


1. Artificial Neural Networks (ANN) consist of layers of interconnected neurons. The
components include input layer, hidden layers, and output layer. Activation functions
introduce non-linearity, allowing networks to learn complex patterns.

2. A CNN consists of convolutional layers, pooling layers, and fully connected layers. The
convolution operation extracts features by applying filters to detect patterns like edges,
textures, and shapes.

3. RNNs maintain memory of previous inputs, making them effective for sequential data like
time series and speech recognition. Traditional neural networks lack this ability.

4. Dropout is a regularization technique that randomly drops neurons during training,


reducing overfitting and improving model generalization.

5. Overfitting occurs when a model memorizes training data but fails on new data. It can be
prevented using dropout, regularization, or increasing training data.

6. Backpropagation is an algorithm used to update network weights by computing gradients


and adjusting parameters to minimize error.
Part C: Long Answer Solutions
1. Supervised learning involves labeled data (e.g., spam detection), unsupervised learning
identifies patterns in unlabeled data (e.g., clustering), and reinforcement learning optimizes
decisions through rewards (e.g., robotics).

2. CNNs use convolutional layers to extract features, pooling layers for downsampling, and
fully connected layers for classification. Filters detect patterns like edges and shapes.

3. Overfitting happens when a model performs well on training but poorly on new data.
Identified by high training accuracy but low validation accuracy. Solutions include dropout,
regularization, and more training data.

You might also like