NEW
Font size
WorksheetsQUIZ
Total questions: 45
Worksheet time: 4mins
What is the focus of this Course?
To introduce machine learning concepts
To cover advanced mathematics
To discuss only deep learning
To focus on data visualization
What method is widely employed in machine learning?
Logistic regression
Support vector machines
Decision trees
K-means clustering
What is discussed in relation to deep networks?
Minimization problem of a mathematical function
Data preprocessing techniques
Feature selection methods
Model evaluation metrics
What technique is used to achieve the mathematical goal?
Gradient descent
Random search
Grid search
Bayesian optimization
What is the focus of this week's discussion?
Model training and transfer learning
Data cleaning and preprocessing
Feature engineering techniques
Hyperparameter tuning
What concept is fundamental in natural language processing?
Word embeddings
Image recognition
Time series forecasting
Clustering algorithms
What is the purpose of the Transformer Network?
To be more flexible and robust than RNN
To simplify data processing
To enhance image classification
To reduce computational costs
What is Reinforcement Learning concerned with?
Maximizing rewards in a situation
Minimizing errors in predictions
Optimizing data storage
Improving user experience
What will be discussed after learning the initial steps of Reinforcement Learning?
Q Learning and Deep Q Learning
Supervised learning techniques
Unsupervised learning methods
Data visualization tools
What is the primary goal of machine learning?
To memorize patterns in data
To make predictions or decisions based on data
To create a new programming language
To replace human intelligence
Which of the following is a supervised learning algorithm?
K-Means
Logistic Regression
Apriori
DBSCAN
In logistic regression, the output is:
A continuous value
A discrete class label
A cluster assignment
A probability distribution
What is the role of the activation function in a neural network?
It adds non-linearity to the model
It calculates the loss function
It normalizes the input features
It optimizes the learning rate
Which activation function is commonly used in deep learning
Sigmoid
ReLU
Tanh
Softmax
What does MLP stand for in neural networks
Multi-layered Processor
Multi-Layer Perceptron
Multi-level Processing
Machine Learning Program
The main advantage of deep learning over traditional ML models is
It requires less data
It can automatically extract features
It does not require hyperparameter tuning
It runs faster than all other models
Which of the following is NOT a deep learning framework?
TensorFlow
PyTorch
Scikit-learn
Keras
What is the primary function of backpropagation in neural networks?
To initialize weights
To update weights based on error gradients
To split data into training and testing sets
To prevent overfitting
Which of these is a commonly used optimizer in deep learning?
Gradient Descent
Stochastic Gradient Descent
K-Means
Principal Component Analysis
CNNs are primarily used for
Time-series forecasting
Image processing
Image processing
Text summarization
What is the role of pooling layers in CNNs?
To reduce dimensionality
To detect edges
To normalize data
To improve gradient flow
Which of these pooling methods is commonly used?
Max pooling
Min pooling
Random pooling
Median pooling
A CNN uses filters to
Reduce overfitting
Extract features from images
Increase image resolution
Detect noise
Transfer learning in CNNs refers to
Using a pre-trained model on a new task
Training from scratch
Replacing layers in the model
Fine-tuning hyperparameters
Word embeddings are used to:
Represent words as numerical vectors
Create decision trees
Normalize text data
Translate between languages
Which algorithm is commonly used to generate word embeddings
Naïve Bayes
PCA
Word2Vec
K-Means
What is an RNN primarily used for?
Image classification
Clustering
Sequential data processing
Feature extraction
LSTMs solve which major issue in RNNs?
Slow computation
Vanishing gradients
Lack of parallel processing
Overfitting
Attention mechanisms help in:
Improving learning rate
Assigning importance to input tokens
Reducing overfitting
Removing stopwords
The Transformer model is primarily used in:
Image processing
Text-based tasks
Reinforcement learning
Anomaly detection
What is the key innovation in Transformers?
Convolutional layers
Self-attention mechanism
Principal component analysis
Decision trees
Transformers replace which traditional neural network structure?
CNNs
RNN
GAN
SVMs
The term "sequence-to-sequence" refers to
Converting structured data into images
Mapping an input sequence to an output sequence
Stacking multiple neural networks
Optimizing loss functions
BERT is an example of:
A CNN model
A Transformer-based model
A reinforcement learning algorithm
A clustering method
Reinforcement learning is based on:
Supervised learning
Maximizing cumulative rewards
Feature selection
Naïve Bayes
The key difference between reinforcement learning and supervised learning is:
RL does not require labeled data
RL uses neural networks
RL is slower
RL cannot generalize
Q-learning is an example of:
Supervised learning
Unsupervised learning
Reinforcement learning
Semi-supervised learning
In reinforcement learning, "exploration" refers to
Selecting actions based on past rewards
Trying new actions
Increasing network depth
Fine-tuning hyperparameters
A deep Q-network (DQN) is used for:
Combining deep learning with Q-learning
Image classification
Clustering data
Feature extraction
Which of the following statements about bias and variance tradeoff is true?
Increasing bias reduces variance, and vice versa
High bias leads to overfitting
High variance leads to underfitting
In a high-dimensional dataset, which technique is best for avoiding the curse of dimensionality?
Increasing the number of training samples
Reducing the number of features using PCA
Using a more complex model
Which optimization algorithm is most effective when dealing with sparse data and large-scale datasets
Stochastic Gradient Descent (SGD)
Batch Gradient Descent
Newton’s Method
What is the main advantage of using L1 regularization (Lasso) over L2 regularization (Ridge)?
L1 leads to sparse feature selection
L1 is computationally more efficient
L1 reduces overfitting better than L2
L1 does not affect feature weights
The Kernel Trick in SVMs is used to:
Transform non-linearly separable data into a higher-dimensional space
Improve the computational efficiency of linear regression
Reduce model complexity
Minimize bias in the dataset
The No Free Lunch Theorem in ML states that:
Neural networks always outperform traditional ML models
No model is best for all datasets
Deep learning works best with small datasets
There exists an optimal universal model
