Can you explain how the choice of kernel in SVM affects the decision boundary?

  • Changes the color
  • Increases computation time
  • Reduces data size
  • Transforms the feature space
Different kernels transform the feature space in different ways, affecting the shape and position of the decision boundary in SVM.

Clustering is a common task in __________ learning, where data is grouped based on inherent similarities without the use of labels.

  • reinforcement
  • semi-supervised
  • supervised
  • unsupervised
Unsupervised learning commonly involves clustering, where data is grouped based on similarities without using labels.

What is the primary difference between the Gini Index and entropy when used in Decision Trees?

  • Calculation Method
  • Complexity
  • Scale
  • Units
Gini Index and entropy are both used to measure purity, but they are calculated differently. Entropy uses logarithms, while Gini Index does not.

When a Decision Tree is too complex and fits the training data too well, __________ techniques can be applied to simplify the model.

  • Bagging
  • Boosting
  • Normalizing
  • Pruning
When a Decision Tree is overfitting (too complex), pruning techniques can be applied to simplify the model. Pruning involves removing branches that have little predictive power, thereby reducing the complexity and the risk of overfitting.

What are the potential issues with using a large value of K in KNN, and how can they be mitigated?

  • All of the above
  • Complexity, can be mitigated by feature selection
  • Overfitting, can be mitigated by reducing dimensions
  • Underfitting, can be mitigated by reducing K
A large value of K in KNN may cause underfitting. It can be mitigated by reducing K or using techniques like cross-validation.

In a situation where the MAE is significantly lower than the RMSE, what might this tell you about the distribution of the errors in your model?

  • Errors are normally distributed; no impact on model
  • Errors are uniformly distributed; no large outliers
  • Many large errors, few small outliers
  • Many small errors, few large outliers
When the Mean Absolute Error (MAE) is significantly lower than the Root Mean Squared Error (RMSE), it likely indicates that the model has many small errors and a few large outliers. RMSE, being sensitive to larger errors, would be higher, while MAE would be less impacted by those larger errors. An analysis of the residuals can further elucidate the nature of these errors.

In the context of SVM, the support vectors are the data points that lie closest to the _________.

  • Data center
  • Hyperplane
  • Kernel
  • Margin
Support vectors are the data points that lie closest to the hyperplane and influence its position and orientation.

Linear Discriminant Analysis (LDA) is often used for dimensionality reduction before applying a classification algorithm, as it seeks to find the axis that best separates the ___________.

  • classes
  • data
  • features
  • variables
LDA seeks to find the axis that "best separates the classes" to reduce dimensionality while retaining class separation.

A company wants to classify its products into different categories based on various features. How could LDA be applied here, considering both within-class and between-class variances?

  • Apply LDA to balance within-class and between-class variances for effective classification
  • Focus on within-class variance and ignore between-class variance
  • Ignore within-class variance and focus on between-class variance
  • Use another method
LDA could be applied by considering both within-class and between-class variances, seeking to "balance within-class and between-class variances for effective classification." This ensures that products in the same category are similar, while products in different categories are distinct.

How does the K-Means clustering algorithm determine the centroids?

  • Based on Density
  • By Class Labels
  • Minimizing Euclidean Distance
  • Random Selection
K-Means clustering algorithm determines the centroids by iteratively minimizing the sum of squared Euclidean distances between the data points and the centroids of their respective clusters.