MCQs

Disclaimer:

Educational Use Only:

The website links provided are to be used strictly for educational and knowledge-gaining purposes only.

No Misuse:

Do not misuse the links or the information contained therein for any commercial, illegal, or unethical activities.

Respect for Websites and Businesses:

Respect the intellectual property and business interests of the websites and their owners. Do not attempt to alter, copy, or redistribute content without proper authorization.

Acknowledge Value of Knowledge:

Recognize and value the knowledge and resources shared by these websites. Use the information responsibly and ethically.

No Endorsement:

The provision of these links does not imply endorsement of the linked sites, their content, or their products and services. Users are responsible for evaluating the suitability and accuracy of the information.

User Responsibility:

By accessing these links, you agree to use the information at your own risk and to comply with all applicable laws and website terms of use.

Overall MCQs for Reference

What is model training in machine learning?
a. Sleeping of model
b. Updating model parameters with data
c. Ignoring input data
d. Deleting parameters
Answer: b. Updating model parameters with datasanfoundry+2
Which of the following is a supervised learning algorithm?
a. K-Means
b. Linear Regression
c. PCA
d. Apriori
Answer: b. Linear Regressionplacementpreparation+2
What is ‘overfitting’ in model training?
a. Model fits training data too well
b. Model ignores test data
c. Model forgets training data
d. Model stops learning
Answer: a. Model fits training data too wellfutureskillsacademy+1
To prevent overfitting, which technique is often used in neural networks?
a. Dropout
b. Adding layers
c. Removing bias
d. None of above
Answer: a. Dropoutgyansetu+1
Which term describes a model that performs well on training data but poorly on testing data?
a. Underfitting
b. Balanced
c. Overfitting
d. Regularized
Answer: c. Overfittingplacementpreparation+1
What does ‘K’ in KNN stand for?
a. Kernel
b. Knowledge
c. Number of neighbors
d. K-means
Answer: c. Number of neighborsyoutubegyansetu
What does epoch mean during training?
a. Training stopped
b. One complete pass through dataset
c. Accuracy calculation
d. Metric evaluation
Answer: b. One complete pass through datasetsanfoundry+1
Which method splits the dataset for model evaluation?
a. validation_split()
b. train_test_split()
c. split_data()
d. random_split()
Answer: b. train_test_split()scribd+1
What is Gradient Descent?
a. Optimization algorithm
b. Data transformation technique
c. Data splitting method
d. Model evaluation metric
Answer: a. Optimization algorithminterviewkickstart+2
Which algorithm uses labeled data for training?
a. K-Means
b. Linear Regression
c. PCA
d. DBSCAN
Answer: b. Linear Regressioninterviewbit+1
Mini-Batch Gradient Descent involves:
a. Training using all data at once
b. Training with batches
c. Training on single data points
d. Ignoring batches
Answer: b. Training with batchesinterviewkickstart+1
Underfitting occurs when:
a. Model captures patterns in training data
b. Model ignores noise
c. Model fails to capture patterns
d. Model memorizes data
Answer: c. Model fails to capture patternsfutureskillsacademy+1
Early stopping is an approach to:
a. Start training
b. Avoid overfitting
c. Increase bias
d. None of these
Answer: b. Avoid overfittinggyansetu+1
Overfitting can be detected by:
a. High accuracy on training set, low on test set
b. High accuracy everywhere
c. Low accuracy everywhere
d. Balanced loss
Answer: a. High accuracy on training set, low on test setplacementpreparation+1
Regularization in model training helps by:
a. Reducing variance
b. Increasing variance
c. Deleting model
d. Adding noise
Answer: a. Reducing variancesanfoundry+2
Which is a regularization technique in linear models?
a. L1
b. L5
c. L0
d. L7
Answer: a. L1sanfoundry+1
Cross-validation is used for:
a. Model accuracy in deployment
b. Model evaluation on different data splits
c. Skipping validation set
d. Increasing bias
Answer: b. Model evaluation on different data splitsinterviewkickstart+2
Batch gradient descent works by:
a. Processing all samples together
b. Processing one sample at a time
c. Processing random batch
d. Ignoring samples
Answer: a. Processing all samples togethersanfoundry
In logistic regression, the cost function commonly used is:
a. Mean squared error
b. Cross Entropy
c. R-squared
d. Root mean squared error
Answer: b. Cross Entropyinterviewkickstart+1
Which type of learning uses both labeled and unlabeled data?
a. Supervised
b. Unsupervised
c. Reinforcement
d. Semi-supervised
Answer: d. Semi-supervisedinterviewbit+1
Hyperparameter tuning is done to:
a. Change model parameters after training
b. Optimize model performance
c. Increase randomness
d. Skip training
Answer: b. Optimize model performancegyansetu+1
Which optimizer adapts the learning rate for each parameter?
a. SGD
b. Linear Regression
c. Adam
d. Decision Tree
Answer: c. Adamgyansetu+1
Model generalization means:
a. Performing well on new, unseen data
b. Memorizing training data
c. Ignoring new data
d. Skipping training
Answer: a. Performing well on new, unseen datagyansetu+1
Ensemble models combine:
a. One model
b. Multiple models
c. Data sources
d. None
Answer: b. Multiple modelsyoutubesanfoundry+1
Bagging is:
a. Boosting variants
b. Bootstrap aggregating
c. Random data removal
d. A clustering algorithm
Answer: b. Bootstrap aggregatingsanfoundry+1
Which metric is usually monitored during model training for best epoch selection?
a. Training loss
b. Validation loss
c. Accuracy
d. Learning rate
Answer: b. Validation lossinterviewkickstart+1
Which of these is a boosting algorithm?
a. Bagging
b. AdaBoost
c. KMeans
d. Decision Tree
Answer: b. AdaBoostgyansetu+1
Which of the following is not a type of learning?
a. Supervised
b. Unsupervised
c. Reinforcement
d. Clustering-only
Answer: d. Clustering-onlydiptisurvase.wordpress+1
Which task is best for unsupervised learning?
a. Classification
b. Regression
c. Clustering
d. Prediction
Answer: c. Clusteringinterviewbit+1
Decision Tree pruning helps:
a. Avoid overfitting
b. Increase depth
c. Add bias
d. Randomize tree
Answer: a. Avoid overfittingdiptisurvase.wordpress+2
Stratified sampling during train-test split ensures:
a. Random selection
b. Balanced class proportions
c. Increased accuracy
d. Smaller dataset
Answer: b. Balanced class proportionsinterviewkickstart+1
Dropout in neural networks:
a. Sets random neurons to zero
b. Adds new layers
c. Removes data instances
d. Increases bias
Answer: a. Sets random neurons to zerosanfoundry+1
In SVM, what does the kernel trick do?
a. Transforms features to higher dimensional space
b. Scales features
c. Deletes features
d. Reduces dimensions
Answer: a. Transforms features to higher dimensional spacegyansetu+1
Stochastic Gradient Descent uses:
a. All data at once
b. Only training set
c. Mini-batches or one sample at a time
d. No data
Answer: c. Mini-batches or one sample at a timeinterviewkickstart+1
Overfitting can be reduced using:
a. Increase tree depth
b. Regularization
c. Batch size increase
d. High learning rate
Answer: b. Regularizationplacementpreparation+2
Which is an unsupervised learning algorithm?
a. K-Means Clustering
b. Decision Trees
c. Random Forest
d. Linear Regression
Answer: a. K-Means Clusteringscribd+1
Which machine learning paradigm maximizes a reward signal?
a. Supervised
b. Unsupervised
c. Reinforcement
d. Regression
Answer: c. Reinforcementfutureskillsacademy+1
Feature scaling benefits model training by:
a. Speeding up optimization
b. Adding bias
c. Increasing model depth
d. Adding more data
Answer: a. Speeding up optimizationplacementpreparation+1
Which of the following is not a hyperparameter?
a. Learning rate
b. Batch size
c. Weight initialization
d. Test accuracy
Answer: d. Test accuracyplacementpreparation+1
Ensemble learning improves results by:
a. Using single tree
b. Averaging predictions from many models
c. Removing weak models
d. Ignoring test set
Answer: b. Averaging predictions from many modelssanfoundry+2
Cross-entropy is commonly used for:
a. Regression loss
b. Classification loss
c. Outlier detection
d. Clustering metric
Answer: b. Classification lossinterviewkickstart+1
What is one-hot encoding useful for?
a. Numeric feature scaling
b. Categorical variable conversion
c. Adding regularization
d. Removing features
Answer: b. Categorical variable conversionplacementpreparation+1
Which optimizer uses momentum to accelerate learning?
a. SGD
b. Adam
c. RMSprop
d. Momentum SGD
Answer: d. Momentum SGDgyansetu+1
Model bias refers to:
a. Error due to assumptions in model
b. Error due to data randomness
c. Error from feature scaling
d. Training error only
Answer: a. Error due to assumptions in modelfutureskillsacademy+1
Data augmentation is used to:
a. Increase dataset diversity
b. Reduce model size
c. Improve test loss
d. Remove features
Answer: a. Increase dataset diversitysanfoundry+1
PCA is used for:
a. Feature selection
b. Feature extraction/dimensionality reduction
c. Model evaluation
d. Test data transformation
Answer: b. Feature extraction/dimensionality reductionscribd+1
What is transfer learning?
a. Training from scratch
b. Using pre-trained model and adapting
c. Shuffling training data
d. Regularization technique
Answer: b. Using pre-trained model and adaptinginterviewkickstart+1
During training, shuffling data helps:
a. Avoid model memorizing order
b. Increase overfitting
c. Reduce accuracy
d. Add bias
Answer: a. Avoid model memorizing ordergyansetu+1
Trainable parameters are:
a. Updated during training
b. Fixed in model
c. Removed after training
d. Not used in prediction
Answer: a. Updated during traininggyansetu+1
Validation set is used to:
a. Evaluate model during training
b. Train model
c. Shuffle data
d. Ignore test loss
Answer: a. Evaluate model during trainingplacementpreparation+2

Blog

Systemethic Technologies

Leave a Reply Cancel reply

Useful links

Other Terms & Policies

Contact Info