Practical 6 - Classification

What is the primary goal of the MNIST dataset?

• To classify images of animals

• To recognize handwritten digits

• To identify types of plants

• To predict weather patterns

How many features does each image in the MNIST dataset have?

• 784

• 1024

• 500

• 256

Which method is used to split the MNIST dataset into training and test sets?

• Random Split

• Temporal Split

• Fixed Split (First 60,000 for training, last 10,000 for testing)

• K-Fold Cross-Validation

Which metric is used to evaluate the performance of the classification models?

• Mean Absolute Error

• Cross-Validation Accuracy

• Root Mean Squared Error

• Adjusted R-Squared

What is a confusion matrix used for in classification tasks?

• To visualize the decision boundaries

• To summarize prediction results and identify misclassifications

• To calculate feature importance

• To preprocess the dataset

What is the purpose of using StandardScaler in the practical session?

• To perform feature selection

• To normalize data for improved performance

• To reduce the number of features

• To balance the dataset

Which performance metric is calculated as the harmonic mean of precision and recall?

• Accuracy

• F1-Score

• Precision-Recall AUC

• ROC AUC

What does the predict_proba method return in the context of classification?

• The predicted class labels

• The probability estimates for each class

• The confusion matrix

• The feature importances

In multiclass classification with SVM, what does the probability=True parameter do?

• Enables the model to perform linear classification

• Allows the model to output probability estimates for each class

• Reduces the dimensionality of the data

• Improves model training speed

What is the main advantage of using cross-validation in evaluating classifiers?

• It provides a single train-test split

• It ensures that the model performs well on unseen data

• It reduces the size of the dataset

• It simplifies model training