Why is underfitting also considered an undesirable property in a machine learning model?
- It enhances generalization
- It fails to capture underlying patterns
- It increases model complexity
- It reduces model bias
Underfitting is undesirable because it fails to capture the underlying patterns in the training data, leading to poor performance on both training and unseen data.
In _______-fold Cross-Validation, each observation is left out once as the validation set, providing a robust estimate of model performance.
- 1
- 2
- k
- n
In 1-fold Cross-Validation, also known as Leave-One-Out Cross-Validation (LOOCV), each observation is left out once as the validation set. It provides a robust estimate of model performance but can be computationally expensive as it requires fitting the model n times, where n is the number of observations in the dataset.
What are the hardware requirements for Deep Learning compared to conventional Machine Learning algorithms, and why is there a difference?
- Less for Deep Learning due to simpler models
- More for Deep Learning due to more complex models and parallel processing
- More for Machine Learning due to more data requirements
- Similar, as they both require the same computational resources
Deep Learning typically requires more hardware resources, such as GPUs, due to the complexity of the models and the need for parallel processing.
In a situation where interpretability is crucial, how would you approach using a Random Forest or Gradient Boosting model?
- Avoid using them entirely
- Provide feature importance scores
- Use without any explanation
- Utilize simpler base learners
In scenarios where interpretability is vital, providing feature importance scores can give insights into the contribution of each feature to the model's predictions. This approach adds an element of transparency to models like Random Forest or Gradient Boosting, which are typically considered more complex.
To handle non-linear patterns in the data, SVM uses the Kernel Trick to transform the data into a _________ space.
- Fixed
- Linear
- Non-linear
- Unchanged
The Kernel Trick can transform the data into a non-linear space, allowing for the classification of non-linear patterns.
Which type of Machine Learning primarily uses classification techniques?
- Reinforcement Learning
- Semi-supervised Learning
- Supervised Learning
- Unsupervised Learning
Supervised learning primarily uses classification techniques as it works with labeled data, allowing the model to learn to predict discrete categories or classes.
What does the first principal component in PCA represent?
- The direction of maximum variance
- The direction of minimum variance
- The least amount of variance in the data
- The mean of the data
The first principal component in PCA represents the direction of maximum variance in the data. It's the line (or hyperplane in higher dimensions) that best captures the structure of the data by explaining the most variance.
The method of ___________ focuses on finding the linear combinations of variables that best separate different classes, making it useful in classification problems.
- Linear Discriminant Analysis
- clustering
- normalization
- scaling
Linear Discriminant Analysis (LDA) focuses on finding the linear combinations of features that best separate different classes. It's especially useful in classification problems where the goal is to distinguish between different categories or groups.
How would you use dimensionality reduction to help visualize a complex, high-dimensional dataset?
- Use PCA to reduce to 2 or 3 dimensions
- Increase the number of dimensions for clarity
- Visualize each feature separately
- Apply clustering first
Using PCA to reduce the data to 2 or 3 dimensions is an effective way to visualize complex, high-dimensional datasets. This transformation retains the most significant patterns while making it possible to plot the data in a 2D or 3D space, thus facilitating the understanding of the underlying structure. Other options do not directly contribute to meaningful visualizations of high-dimensional data.
You have implemented K-Means clustering but are getting inconsistent results. What could be the reason related to centroid initialization?
- Centroids initialized with zero values
- Centroids too close to each other
- Random initialization leading to different results
- Too many centroids
Random initialization of centroids in K-Means can lead to inconsistent results across different runs, as the initial positioning of centroids can affect the final cluster formation.