DSE DSE - Data Science Deep Learning and Neural Networks 2 — Questions and Answers
Question 1: Which neural network architecture processes sequential data by maintaining a hidden state that captures information from previous time steps?
- CNN
- Autoencoder
- Recurrent Neural Network (RNN) (Correct answer)
- Multilayer Perceptron
Correct answer: Recurrent Neural Network (RNN)
RNNs have recurrent connections that pass the hidden state from one time step to the next, enabling them to model temporal dependencies in sequential data like text or time series.
Question 2: What is transfer learning in deep learning?
- Moving a trained model to a different hardware device
- Reusing a model pretrained on one task as the starting point for training on a related task (Correct answer)
- Converting a model from one framework to another
- Transferring training data between datasets
Correct answer: Reusing a model pretrained on one task as the starting point for training on a related task
Transfer learning leverages features learned by a model on a large source task (e.g., ImageNet) and fine-tunes it on a smaller target task, dramatically reducing data and compute requirements.
Question 3: What does batch normalization do in a neural network?
- Randomly drops neurons to prevent overfitting
- Normalizes the inputs to each layer to have zero mean and unit variance within each mini-batch (Correct answer)
- Increases the learning rate automatically
- Reduces the number of parameters in the model
Correct answer: Normalizes the inputs to each layer to have zero mean and unit variance within each mini-batch
Batch normalization standardizes layer inputs within each mini-batch, stabilizing training, reducing sensitivity to weight initialization, and allowing higher learning rates.
Question 4: What is the primary function of a pooling layer in a Convolutional Neural Network?
- Applying a non-linear activation function
- Adding more learnable parameters
- Downsampling feature maps to reduce spatial dimensions and computation (Correct answer)
- Connecting all neurons to the output layer
Correct answer: Downsampling feature maps to reduce spatial dimensions and computation
Pooling layers (e.g., max pooling) reduce the height and width of feature maps by aggregating values in local regions, decreasing computation and providing some translation invariance.
Question 5: What is an autoencoder primarily used for in deep learning?
- Supervised image classification
- Unsupervised dimensionality reduction and feature learning by learning to compress and reconstruct data (Correct answer)
- Training reinforcement learning agents
- Generating adversarial examples
Correct answer: Unsupervised dimensionality reduction and feature learning by learning to compress and reconstruct data
An autoencoder trains an encoder to compress input into a lower-dimensional latent representation and a decoder to reconstruct the original input, learning compact data representations without labels.
Question 6: Which loss function is typically used for binary classification problems in neural networks?
- Mean squared error
- Categorical cross-entropy
- Binary cross-entropy (Correct answer)
- Hinge loss
Correct answer: Binary cross-entropy
Binary cross-entropy measures the dissimilarity between the predicted probability and the true binary label, penalizing confident wrong predictions most heavily.
DSE DSE - Data Science Deep Learning and Neural Networks 2
Which neural network architecture processes sequential data by maintaining a hidden state that captures information from previous time steps?