In a case where you have a dataset with numerous outliers, which clustering algorithm would you choose and why?
- DBSCAN due to robustness to outliers
- DBSCAN due to sensitivity to noise
- K-Means due to robustness to noise
- K-Means due to sensitivity to outliers
DBSCAN (Density-Based Spatial Clustering of Applications with Noise) would be suitable since it's robust to outliers. It can identify dense clusters and leave outliers as unclassified, making it effective in such scenarios.
How can you detect whether a model is overfitting or underfitting the data?
- By analyzing the training and validation errors
- By increasing model complexity
- By looking at the model's visualizations
- By reducing model complexity
Detecting overfitting or underfitting can be done "by analyzing the training and validation errors." Overfitting shows high training accuracy but low validation accuracy, while underfitting shows poor performance on both.
Describe a scenario where you would use the F1-Score as the main performance metric, and explain why it would be suitable.
- In a balanced dataset, to ensure model fairness
- In a scenario where only false negatives are important
- In an imbalanced dataset, to balance both false positives and false negatives
- nan
F1-Score is especially suitable for imbalanced datasets, as it balances both Precision and Recall, ensuring that the model does not bias towards the majority class. It gives an equal weight to false positives and false negatives, providing a more holistic evaluation of the model's performance.
What is the branch of Machine Learning that involves neural networks with three or more layers, which work to analyze various factors of data?
- Deep Learning
- Reinforcement Learning
- Supervised Learning
- Unsupervised Learning
Deep Learning is a subset of Machine Learning that uses neural networks with three or more layers to analyze complex patterns in data.
What are the advantages and limitations of using Bootstrapping in Machine Learning?
- Fast computation but lacks precision
- Reduced bias but increased computation complexity
- Robust statistical estimates but can introduce high variance
- Robust statistical estimates but may not always be appropriate for all data types
The advantages of Bootstrapping include robust statistical estimates, even with small samples, by resampling with replacement. However, it may not always be appropriate for all data types, especially if the underlying distribution of the data is not well represented by resampling. It provides valuable insights but needs to be applied considering the nature of the data and problem.
The addition of _________ in the loss function is a common technique to regularize the model and prevent overfitting.
- bias
- learning rate
- regularization terms
- weights
Regularization terms (like L1 or L2 penalties) in the loss function constrain the model, reducing the risk of overfitting by preventing large weights.
You have trained an SVM but the decision boundary is not fitting well to the data. How could adjusting the hyperplane parameters help?
- Change the kernel's color
- Increase the size of the hyperplane
- Modify the regularization parameter 'C'
- Reduce the number of support vectors
Adjusting the regularization parameter 'C' controls the trade-off between margin maximization and error minimization, helping to fit the decision boundary better.
Discuss the difference between Euclidean distance and Manhattan distance metrics in the context of KNN.
- Euclidean is faster, Manhattan is more accurate
- Euclidean is for 3D, Manhattan for 2D
- Euclidean is for continuous data, Manhattan for categorical
- Euclidean uses squares, Manhattan uses absolutes
Euclidean distance is the square root of the sum of squared differences, while Manhattan distance is the sum of the absolute differences.
A dataset with very high between-class variance but low within-class variance is given. How would the LDA approach be beneficial here?
- LDA would be the same as PCA
- LDA would perform optimally due to the variance characteristics
- LDA would perform poorly
- LDA would require transformation of the dataset
LDA would "perform optimally" in this scenario, as high between-class variance and low within-class variance align perfectly with its objective of maximizing between-class variance and minimizing within-class variance.
What is the risk of using the same data for both training and testing in a Machine Learning model?
- Increase in accuracy; Reduction in bias
- Increase in complexity; Reduction in training time
- Reduction in training time; Increase in bias
- Risk of overfitting; Unrealistic performance estimates
Using the same data for training and testing leads to the risk of overfitting and provides unrealistic performance estimates. The model will have seen all the data during training, so it might not generalize well to new, unseen instances.