How would you handle a multi-class classification problem with LDA, and what is the importance of maximizing between-class variance in this context?
- Apply LDA separately to each class; no importance of between-class variance
- Apply LDA to all classes; maximize between-class variance for class separability
- Ignore between-class variance and focus on within-class variance
- Use another method entirely
For a multi-class classification problem, you would "apply LDA to all classes" and maximize between-class variance. This is essential for separating the classes from each other, which improves classification performance.
The _________ is a single summary value that illustrates the ability of a classification model to discriminate between positive and negative classes.
- AUC
- F1-Score
- Precision
- Recall
The AUC (Area Under the Curve) provides a single summary value representing the model's ability to differentiate between positive and negative classes, considering various thresholds.
How does the Kernel Trick transform the input data?
- Compresses data
- Maps data into higher-dimensional space
- Normalizes data
- Reduces dimensionality
The Kernel Trick maps the input data into a higher-dimensional space to enable linear separation in SVM.
How do training techniques differ between traditional Machine Learning and Deep Learning?
- Deep Learning relies on neural networks with multiple layers
- Deep Learning uses fewer hyperparameters
- Machine Learning requires more data
- Machine Learning uses more complex algorithms
Deep Learning uses neural networks with multiple layers and various training techniques like backpropagation, while traditional Machine Learning may use different algorithms and methods.
Explain the process of selecting the number of principal components in PCA.
- By choosing an arbitrary number
- By selecting all eigenvectors
- By using only the first eigenvector
- By using the elbow method and the cumulative explained variance
The number of principal components in PCA can be selected by considering the cumulative explained variance and looking for an "elbow" in the plot, where adding more components does not significantly increase the explained variance.
The regularization parameter 'C' in SVM controls the trade-off between maximizing the margin and minimizing the _________.
- Kernel size
- Margin
- Misclassification
- Variance
The 'C' parameter controls the trade-off between maximizing the margin and minimizing misclassification.
If a model performs well on the training data but poorly on the validation data, it may be suffering from _________.
- Generalization
- Optimization
- Overfitting
- Underfitting
Overfitting occurs when a model fits too closely to the training data and captures noise, leading to poor performance on unseen data such as the validation set.
The _________ method in Cross-Validation ensures that each fold maintains the same proportion of different classes as the complete dataset.
- k-fold
- proportional
- repeated
- stratified
The stratified method in Cross-Validation ensures that each fold maintains the same proportion of different classes as the complete dataset. It aims to preserve the percentage of samples for each class, especially important when working with imbalanced class distributions.
You have been asked to develop a model that can classify images of handwritten digits. Describe how you would approach this problem using classification algorithms.
- Analyze images without preprocessing
- Convert images into numerical data; use algorithms like CNN
- Use regression algorithms
- Use time-series analysis
Converting images into numerical data (pixel values) and using deep learning algorithms like Convolutional Neural Networks (CNNs) can be an effective approach for classifying handwritten digits, as CNNs are well-suited for image data.
How does reinforcement learning differ from supervised learning in terms of feedback and learning process?
- Both learn from environment
- Feedback through rewards/penalties
- No feedback involved
- Uses labeled data
Reinforcement learning uses feedback through rewards and penalties to guide the learning process, unlike supervised learning with labeled data.