Explain how Ridge and Lasso handle multicollinearity among the features.
- Both eliminate correlated features
- Both keep correlated features
- Ridge eliminates correlated features; Lasso keeps them
- Ridge keeps correlated features; Lasso eliminates them
Ridge regularization keeps correlated features but shrinks coefficients; Lasso can eliminate some by setting coefficients to zero.
What are some common applications for each of the four types of Machine Learning: Supervised, Unsupervised, Semi-Supervised, and Reinforcement?
- Specific to finance
- Specific to healthcare
- Specific to manufacturing
- Varies based on the problem domain
The applications for these types of Machine Learning vary and can be tailored to various problem domains, not confined to specific industries.
What is the difference between simple linear regression and multiple linear regression?
- Number of dependent variables
- Number of equations
- Number of independent variables
- Number of observations
Simple linear regression involves one independent variable to predict a dependent variable, whereas multiple linear regression uses two or more independent variables for prediction. The inclusion of more variables in multiple linear regression allows for more complex models and can lead to a better understanding of the relationships between variables.
The performance of an LDA model can be evaluated using ___________, which considers both within-class and between-class variances.
- accuracy metrics
- error rate
- feature selection
- principal components
"Accuracy metrics" that consider both within-class and between-class variances can be used to evaluate the performance of an LDA model. It gives a comprehensive view of how well the model has separated the classes.
Name a popular algorithm used in classification problems.
- Clustering
- Decision Trees
- Linear Regression
- Principal Component Analysis
Decision Trees are a popular algorithm used in classification problems. They work by recursively partitioning the data into subsets based on feature values, leading to a decision on the class label.
What is regression in the context of Machine Learning?
- A type of classification
- Analyzing computer performance
- Predicting a categorical value
- Predicting a numerical value
Regression in Machine Learning refers to the process of predicting a continuous numerical outcome or dependent variable based on one or more independent variables. It plays a crucial role in various areas, like economics, where it might be used to predict prices, or in biology, to forecast growth rates.
In KNN, how does an increase in the value of K generally affect the bias and variance of the model?
- Decreases bias, increases variance
- Decreases both bias and variance
- Increases bias, decreases variance
- Increases both bias and variance
Increasing the value of K generally increases bias and decreases variance in the KNN model.
You've trained a model with a small training set and a large testing set. What challenges might you encounter, and how could they be addressed?
- Both Overfitting and Underfitting
- Data is perfectly balanced
- Overfitting
- Underfitting
A small training set might lead to overfitting, where the model memorizes noise from the training data. Conversely, it might also lead to underfitting if the model fails to capture the underlying pattern. Cross-validation, bootstrapping, or augmenting the training set with additional relevant data can help balance the model's ability to generalize.
In Supervised Learning, _________ and ___________ are the two main types of problems.
- Classification; Clustering
- Classification; Regression
- Regression; Clustering
- Regression; Ensemble Learning
In Supervised Learning, the two main types of problems are Classification and Regression. Classification is about categorizing data into predefined classes, while Regression is predicting a continuous outcome.
You've built a multiple linear regression model and found that two or more predictors are highly correlated. What problems might this cause, and how can you solve them?
- High bias, Address by increasing the model complexity
- High variance, Address by using Lasso regression
- Overfitting, Address by removing correlated features or using Ridge regression
- Underfitting, Address by adding more features
Multicollinearity, where predictors are highly correlated, can cause overfitting and unstable estimates. This can be addressed by removing correlated features or using Ridge regression, which penalizes large coefficients and reduces the impact of multicollinearity.