1. . Advanced deep learning focuses on:
(A) Compressing datasets only
(B) Encrypting neural networks
(C) Complex architectures and techniques such as CNNs, RNNs, LSTMs, GANs, and Transformers
(D) Backup only
2. . Convolutional Neural Networks (CNNs) are best suited for:
(A) Backup only
(B) Encrypting images
(C) Compressing image data
(D) Image recognition, object detection, and computer vision tasks
3. . Key layers in CNN include:
(A) Encrypting layers
(B) Convolutional layer, pooling layer, fully connected layer
(C) Compression layers
(D) Backup only
4. . Pooling layers in CNNs are used to:
(A) Encrypt pooled data
(B) Reduce spatial dimensions and computational complexity while retaining important features
(C) Compress features
(D) Backup only
5. . Recurrent Neural Networks (RNNs) are suitable for:
(A) Compressing sequences
(B) Encrypting sequences
(C) Sequential data such as text, speech, and time series
(D) Backup only
6. . Long Short-Term Memory (LSTM) networks help to:
(A) Compress sequences
(B) Encrypt memory
(C) Solve the vanishing gradient problem in RNNs and capture long-term dependencies
(D) Backup only
7. . Gated Recurrent Units (GRUs) are:
(A) Compressing GRUs
(B) Encrypting GRUs
(C) Simplified versions of LSTMs with fewer parameters
(D) Backup only
8. . Generative Adversarial Networks (GANs) consist of:
(A) A generator and a discriminator network competing in a zero-sum game
(B) Encrypting GANs
(C) Compressing GANs
(D) Backup only
9. . GANs are mainly used for:
(A) Data generation, image synthesis, and augmentation
(B) Encrypting images
(C) Compressing datasets
(D) Backup only
10. . Autoencoders are used for:
(A) Compressing features only
(B) Encrypting data
(C) Dimensionality reduction, denoising, and unsupervised feature learning
(D) Backup only
11. . Attention mechanism in deep learning helps to:
(A) Encrypt attention weights
(B) Focus on important parts of input sequences for better prediction
(C) Compress sequences
(D) Backup only
12. . Transformers are:
(A) Backup only
(B) Encrypting transformers
(C) Compressing transformers
(D) Deep learning architectures that rely entirely on attention mechanisms, widely used in NLP
13. . BERT (Bidirectional Encoder Representations from Transformers) is used for:
(A) Encrypting text
(B) NLP tasks such as text classification, question answering, and sentiment analysis
(C) Compressing text
(D) Backup only
14. . Transfer learning in deep learning involves:
(A) Backup only
(B) Encrypting weights
(C) Compressing models
(D) Using a pre-trained model on a new but related task to save time and improve performance
15. . Fine-tuning is:
(A) Adjusting weights of a pre-trained model on new data to adapt it to a specific task
(B) Encrypting models
(C) Compressing weights
(D) Backup only
16. . Dropout in advanced deep learning is used to:
(A) Backup only
(B) Encrypt neurons
(C) Compress layers
(D) Prevent overfitting by randomly deactivating neurons during training
17. . Batch normalization helps to:
(A) Encrypt batches
(B) Accelerate training and stabilize learning by normalizing layer inputs
(C) Compress layers
(D) Backup only
18. . Residual connections in deep networks:
(A) Encrypt residuals
(B) Help train very deep networks by allowing gradients to flow directly
(C) Compress layers
(D) Backup only
19. . Hyperparameter tuning in deep learning involves:
(A) Selecting optimal values for learning rate, batch size, number of layers, and neurons
(B) Encrypting hyperparameters
(C) Compressing hyperparameters
(D) Backup only
20. . The main purpose of advanced deep learning is to:
(A) Solve complex tasks in vision, language, and sequential data using sophisticated neural network architectures
(B) Encrypt all models
(C) Compress all features
(D) Backup only