What are the underlying assumptions of LDA, and how do they affect the performance?
- Assumes different covariance matrices, normal distribution; affects adaptability
- Assumes equal class sizes; affects bias
- Assumes equal variance, non-normal distribution; affects robustness
- Assumes normal distribution, equal covariance matrices; affects classification accuracy
LDA assumes that the features are normally distributed and that the classes have equal covariance matrices. These assumptions, if met, lead to better "classification accuracy," but if violated, may lead to suboptimal performance.
__________ is a boosting algorithm that builds trees one at a time, where each tree corrects the errors of the previous one.
- AdaBoost
- Bagging
- Gradient Boosting
- Random Forest
AdaBoost is a boosting algorithm that builds trees sequentially, with each tree correcting the errors of the previous one by giving more weight to the misclassified instances.
What is the relationship between dimensionality reduction and the curse of dimensionality?
- Depends on the method
- Has no relationship
- Helps in mitigating the curse
- Intensifies the curse
The relationship between dimensionality reduction and the curse of dimensionality is that dimensionality reduction helps in mitigating the curse. The curse of dimensionality refers to the challenges that arise as the number of dimensions increases, and reducing dimensions can alleviate these problems.
You need to improve the performance of a weak learner. Which boosting algorithm would you select, and why?
- AdaBoost
- Any boosting algorithm will suffice
- Gradient Boosting without considering the loss function
- Random Boosting
AdaBoost is a boosting algorithm designed to improve the performance of weak learners. By adjusting the weights of misclassified instances and focusing on them in subsequent models, AdaBoost iteratively corrects errors and enhances the overall model's performance.
You are using KNN for a regression problem. What are the special considerations in selecting K and the distance metric, and how would you evaluate the model's performance?
- Choose K and metric considering data characteristics, evaluate using regression metrics
- Choose fixed K and Manhattan metric, evaluate using recall
- Choose large K and any metric, evaluate using accuracy
- Choose small K and Euclidean metric, evaluate using precision
Selecting K and distance metric considering the data characteristics and evaluating the model using regression metrics like RMSE or MAE is the right approach for KNN in regression.
A dataset contains both categorical and numerical features. Which ensemble method might be suitable, and what preprocessing might be required?
- Random Forest with no preprocessing
- Random Forest with normalization
- Random Forest with one-hot encoding
- Random Forest with scaling
Random Forest is an ensemble method suitable for handling both categorical and numerical features. For categorical features, one-hot encoding might be required to convert them into a numerical format that the algorithm can process.
In Decision Trees, the __________ is used to measure the impurity of a data partition or set.
- Accuracy
- Bias
- Gini Index
- Training set
In Decision Trees, the Gini Index is used to measure the impurity or disorder of a data partition or set. A lower Gini Index value indicates a purer node, and it is used to determine the best splits.
What are some advanced techniques to prevent overfitting in a deep learning model?
- Regularization, Dropout, Early Stopping, Data Augmentation
- Regularization, Dropout, Early Stopping, Over-sampling
- Regularization, Dropout, Late Stopping, Data Augmentation
- Regularization, Over-sampling, Early Stopping, Data Reduction
Advanced techniques such as "Regularization, Dropout, Early Stopping, and Data Augmentation" help in preventing overfitting by adding constraints, randomly deactivating neurons, halting training, and expanding the dataset, respectively.
Describe the process of Bootstrapping and its applications in model evaluation.
- Repeated sampling with replacement for bias reduction
- Repeated sampling with replacement for variance reduction
- Repeated sampling with replacement to estimate statistics and evaluate models
- Repeated sampling without replacement for model validation
Bootstrapping involves repeated sampling with replacement to estimate statistics and evaluate models. By creating numerous "bootstrap samples," it allows the calculation of standard errors, confidence intervals, and other statistical properties, even with a small dataset. It's valuable for model evaluation, hypothesis testing, and providing insight into the estimator's distribution.
You're comparing two Polynomial Regression models: one with a low degree and one with a high degree. The higher degree model fits the training data perfectly but has poor test performance. How do you interpret this, and what actions would you take?
- Choose the high degree model
- Choose the low degree model or consider regularization
- Ignore test performance
- Increase the degree further
The high degree model is likely overfitting the training data, leading to poor test performance. Choosing the low degree model or applying regularization to the high degree model can improve generalization.