2025/2026 CAIP DOMAIN 3 EXAM
QUESTIONS AND DETAILED CORRECT
ANSWERS | A+ GRADE VERIFIED
ANSWERS
What is t-SNE commonly used for? Correct Answer For
visualizing high-dimensional data in 2D or 3D space.
Why is feature scaling important for certain machine
learning algorithms? Correct Answer To ensure that all
features contribute equally to distance-based algorithms.
What are the key benefits of standardization? Correct
Answer It improves the performance of models like SVM
and K-Means by ensuring consistent feature scales.
How does normalization differ from standardization?
Correct Answer Normalization scales features to a fixed
range, while standardization transforms features to have a
mean of 0 and a standard deviation of 1.
What is multicollinearity, and how does it affect models?
Correct Answer Multicollinearity occurs when features are
highly correlated, leading to model instability.
How can multicollinearity be detected in datasets? Correct
Answer By using correlation matrices or Variance Inflation
Factor (VIF) scores.
,Why is recursive feature elimination (RFE) used in feature
selection? Correct Answer RFE iteratively removes the
least important features to improve model accuracy.
What are the risks of feature redundancy in machine
learning models? Correct Answer It increases model
complexity without improving performance.
What is the purpose of regularization in machine learning?
Correct Answer To prevent overfitting and improve the
generalization of the model.
How does L1 regularization (Lasso) work? Correct Answer
It applies a penalty equal to the absolute value of
coefficients, leading to sparse solutions.
What is the key benefit of L2 regularization (Ridge)?
Correct Answer It penalizes large coefficients and reduces
their impact without setting them to zero.
When should Elastic Net be used for regularization?
Correct Answer When there are many correlated features
and both feature selection and regularization are needed.
What is the role of dropout in deep learning models?
Correct Answer Dropout randomly removes neurons
during training, preventing overfitting.
How does batch normalization stabilize the training
process? Correct Answer Batch normalization normalizes
,the input to each layer, improving training stability and
convergence.
What is the main trade-off when using dropout in neural
networks? Correct Answer It increases training time since
the model has to learn with fewer neurons during each
iteration.
How does L1 regularization help in feature selection?
Correct Answer It selects the most important features by
shrinking irrelevant coefficients to zero.
What is the benefit of combining L1 and L2 regularization
(Elastic Net)? Correct Answer It balances feature selection
with regularization strength, improving model
performance.
Why is batch normalization important in deep networks?
Correct Answer It prevents gradients from vanishing or
exploding by stabilizing inputs.
What is convergence in machine learning? Correct
Answer Convergence refers to the process of a model
reaching its optimal solution during training.
How does the learning rate affect model convergence?
Correct Answer The learning rate controls the size of the
steps taken during training; a high rate speeds up
convergence but risks overshooting.
, What is the role of adaptive learning rates like Adam in
optimization? Correct Answer Adam adjusts learning rates
dynamically, making it effective for deep neural networks.
How does early stopping prevent overfitting? Correct
Answer It stops training when performance on the
validation set begins to degrade, preventing overfitting.
What is gradient clipping, and when is it used? Correct
Answer It prevents gradients from becoming too large and
destabilizing training.
How do vanishing gradients affect deep learning models?
Correct Answer Vanishing gradients cause small updates
during backpropagation, making training slow or
ineffective.
What are exploding gradients, and how do they impact
training? Correct Answer Exploding gradients cause
model weights to grow too large, destabilizing training.
How does gradient clipping help stabilize training? Correct
Answer It caps gradients at a predefined value, ensuring
stable updates.
Why is early stopping often used with dropout in deep
learning? Correct Answer It helps prevent overfitting by
stopping training at the optimal point.
QUESTIONS AND DETAILED CORRECT
ANSWERS | A+ GRADE VERIFIED
ANSWERS
What is t-SNE commonly used for? Correct Answer For
visualizing high-dimensional data in 2D or 3D space.
Why is feature scaling important for certain machine
learning algorithms? Correct Answer To ensure that all
features contribute equally to distance-based algorithms.
What are the key benefits of standardization? Correct
Answer It improves the performance of models like SVM
and K-Means by ensuring consistent feature scales.
How does normalization differ from standardization?
Correct Answer Normalization scales features to a fixed
range, while standardization transforms features to have a
mean of 0 and a standard deviation of 1.
What is multicollinearity, and how does it affect models?
Correct Answer Multicollinearity occurs when features are
highly correlated, leading to model instability.
How can multicollinearity be detected in datasets? Correct
Answer By using correlation matrices or Variance Inflation
Factor (VIF) scores.
,Why is recursive feature elimination (RFE) used in feature
selection? Correct Answer RFE iteratively removes the
least important features to improve model accuracy.
What are the risks of feature redundancy in machine
learning models? Correct Answer It increases model
complexity without improving performance.
What is the purpose of regularization in machine learning?
Correct Answer To prevent overfitting and improve the
generalization of the model.
How does L1 regularization (Lasso) work? Correct Answer
It applies a penalty equal to the absolute value of
coefficients, leading to sparse solutions.
What is the key benefit of L2 regularization (Ridge)?
Correct Answer It penalizes large coefficients and reduces
their impact without setting them to zero.
When should Elastic Net be used for regularization?
Correct Answer When there are many correlated features
and both feature selection and regularization are needed.
What is the role of dropout in deep learning models?
Correct Answer Dropout randomly removes neurons
during training, preventing overfitting.
How does batch normalization stabilize the training
process? Correct Answer Batch normalization normalizes
,the input to each layer, improving training stability and
convergence.
What is the main trade-off when using dropout in neural
networks? Correct Answer It increases training time since
the model has to learn with fewer neurons during each
iteration.
How does L1 regularization help in feature selection?
Correct Answer It selects the most important features by
shrinking irrelevant coefficients to zero.
What is the benefit of combining L1 and L2 regularization
(Elastic Net)? Correct Answer It balances feature selection
with regularization strength, improving model
performance.
Why is batch normalization important in deep networks?
Correct Answer It prevents gradients from vanishing or
exploding by stabilizing inputs.
What is convergence in machine learning? Correct
Answer Convergence refers to the process of a model
reaching its optimal solution during training.
How does the learning rate affect model convergence?
Correct Answer The learning rate controls the size of the
steps taken during training; a high rate speeds up
convergence but risks overshooting.
, What is the role of adaptive learning rates like Adam in
optimization? Correct Answer Adam adjusts learning rates
dynamically, making it effective for deep neural networks.
How does early stopping prevent overfitting? Correct
Answer It stops training when performance on the
validation set begins to degrade, preventing overfitting.
What is gradient clipping, and when is it used? Correct
Answer It prevents gradients from becoming too large and
destabilizing training.
How do vanishing gradients affect deep learning models?
Correct Answer Vanishing gradients cause small updates
during backpropagation, making training slow or
ineffective.
What are exploding gradients, and how do they impact
training? Correct Answer Exploding gradients cause
model weights to grow too large, destabilizing training.
How does gradient clipping help stabilize training? Correct
Answer It caps gradients at a predefined value, ensuring
stable updates.
Why is early stopping often used with dropout in deep
learning? Correct Answer It helps prevent overfitting by
stopping training at the optimal point.
