NEW
Font size
WorksheetsMachine Learning & Deep Learning Quiz
Total questions: 50
Worksheet time: 25mins
Which of the following is a type of dataset used in Machine Learning?
Training dataset
Temporary dataset
Background dataset
Random dataset
Continuous data refers to:
Data that can only take fixed categories
Data measured on an infinite scale
Data with labels
Data without labels
Which of the following is labeled data?
Collection of tweets without sentiment tags
Images tagged as “cat” or “dog”
Audio recordings with no labels
Random survey responses
The main purpose of a test dataset is to:
Train the model
Validate the hyperparameters
Evaluate model performance
Store unused data
A flowchart for Machine Learning usually begins with:
Model evaluation
Data collection
Feature engineering
Hyperparameter tuning
Supervised learning requires:
Only unlabeled data
Partially labeled data
Fully labeled data
Randomized data
An example of unsupervised learning is:
Linear regression
K-means clustering
Decision trees
Logistic regression
Which of the following is an example of reinforcement learning?
Predicting house prices
A robot learning to walk
Identifying spam emails
Customer segmentation
Semi-supervised learning uses:
Only labeled data
Only unlabeled data
Both labeled and unlabeled data
No data
Which ML model is prone to overfitting if not pruned?
Linear regression
Decision trees
Naïve Bayes
K-means
Which algorithm works best for text classification?
K-means
Naïve Bayes
KNN
DBSCAN
In SVM, the kernel trick is used for:
Feature selection
Handling non-linear data
Overfitting
Feature scaling
Ridge regression uses:
L1 regularization
L2 regularization
Dropout
Bagging
Which method combines multiple decision trees?
Logistic regression
Random forest
Linear regression
PCA
Gradient boosting differs from random forests because:
Uses bagging
Builds trees sequentially
Uses PCA
Is unsupervised
Feature scaling is important because:
Prevents missing values
Keeps all features on comparable scale
Reduces dataset size
Prevents bias in labels
Which library is primarily for data visualization?
NumPy
Matplotlib
TensorFlow
Scikit-learn
Which deep learning library is developed by Google?
PyTorch
TensorFlow
Scikit-learn
OpenCV
Which library is widely used for computer vision tasks?
TensorFlow
OpenCV
NLTK
Statsmodels
Which library is used for NLP?
Pandas
NLTK
OpenCV
PyTorch
Accuracy is:
(TP+TN)/(Total samples)
TP/(TP+FP)
TP/(TP+FN)
2*(Precision*Recall)/(Precision+Recall)
Precision focuses on:
Correctly identifying all positives
Correctly identifying all negatives
Correctness of positive predictions
Misclassification rate
Recall is:
TP/(TP+FN)
TP/(TP+FP)
(TP+TN)/(Total samples)
FP/(FP+TN)
F1-score balances:
Accuracy & Recall
Precision & Recall
Accuracy & Precision
Recall & Loss
Which metric is best for imbalanced datasets?
Accuracy
F1-score
RMSE
R²
RMSE is always:
Less than MSE
Greater than or equal to MSE’s square root
Negative
Equal to R²
R² (coefficient of determination) indicates:
Correlation between predicted & actual values
Error in predictions
False positive rate
Normalization factor
A confusion matrix is used for:
Regression
Classification
Clustering
Dimensionality reduction
ROC-AUC is useful for:
Comparing regression models
Evaluating classification thresholds
Visualizing clustering
Feature engineering
Logloss measures:
Error in classification probabilities
Training speed
Bias in regression
Clustering variance
Clustering metric ARI stands for:
Adjusted Random Index
Adjusted Rand Index
Average Random Index
Artificial Ranking Index
Dimensionality reduction is useful for:
Increasing dataset size
Reducing computational complexity
Labeling data
Increasing accuracy only
Which dimensionality reduction technique uses eigenvectors?
PCA
t-SNE
K-means
DBSCAN
Which unsupervised algorithm detects anomalies?
Logistic regression
DBSCAN
Naïve Bayes
Linear regression
Association rule learning is mainly used in:
Customer segmentation
Market basket analysis
Regression
Clustering
CNNs are mainly used for:
Sequential data
Image processing
Probability estimation
Text summarization
RNNs are best for:
Image classification
Sequential data like time series
Regression
Clustering
Transformers replaced RNNs in NLP because:
They require less memory
They process sequences in parallel
They cannot overfit
They don’t use embeddings
GANs are used for:
Generating new synthetic data
Feature scaling
Regression
Clustering
Autoencoders are used for:
Supervised learning
Feature extraction & dimensionality reduction
Classification only
Data labeling
A diffusion model is mainly used for:
Data cleaning
High-quality image generation
Feature scaling
Regression
Overfitting occurs when:
Model generalizes well
Model memorizes training data
Model under-learns
Model is regularized
Underfitting means:
Model performs poorly on training data
Model performs poorly on test data only
Model has too many parameters
Model generalizes too much
A good fit is when:
Training accuracy >> Test accuracy
Training accuracy ≈ Test accuracy
Training accuracy < Test accuracy
None
Deep reinforcement learning combines:
Clustering & regression
Deep learning & reinforcement learning
CNNs & RNNs
PCA & GANs
ML focuses on:
Manual feature extraction
Automatic feature extraction
Only image processing
Only text
DL focuses on:
Manual feature engineering
Automated feature extraction
Only regression tasks
Dimensionality reduction
Which datatype is discrete?
Temperature in Celsius
Number of students in a class
Height of a person
Weight in kilograms
Which preprocessing method is used for categorical variables?
Standardization
One-hot encoding
Min-max scaling
Normalization
The ML pipeline generally ends with:
Data collection
Model deployment
Feature engineering
Hyperparameter tuning
