Principle Component Analysis Flashcards
Was bedeutet das fΓΌr A?
Wie ist grob die Definition von Eigenvektoren und Eigenwerten?
Mit welcher Formel bestimmt man Eigenvektoren?
Wie bestimmt man die Kovarianzmatrix?
Wie normalisiert man Daten?
Wie funktioniert das Whitening von Daten?
Which of the following statements are true?
1. Correlation implies causation.
2. Negative covariance of two random variables implies that they are not correlated.
3. πππ£(π΄,π΅)>0 means that A and B tend to move in the same direction.
4. πππ£(π΄,π΅)=0 implies statistical independence of the two random variables A and B.
- πππ£(π΄,π΅)>0 means that A and B tend to move in the same direction.
Which of the following statements on normalization are true?
1. Min-max normalization is suitable for very noisy data.
2. Decimal scaling normalization effectively moves the decimal point to standardize the scale of some data.
3. Z-score normalization decorrelates the data.
4. Data whitening transforms the data into a new coordinate system.
- Decimal scaling normalization effectively moves the decimal point to standardize the scale of some data.
- Data whitening transforms the data into a new coordinate system.
Which of the following statements are correct?
1. Univariate feature selection may fail if multiple features are needed to explain certain behavior.
2. Feature extraction refers to the process of choosing the optimal subset of features according to an objective function.
3. Features may be ranked by comparison of their means and variances with respect to different classes, thereby favoring large differences of the means and low variances.
- und 3.
Was ist univariate feature selection?
Univariate feature selection methods evaluate each feature independently, based on its individual contribution to the target variable.
Which of the following statements about sampling are correct?
1. The goal of a good sampling method is to achieve a representative sample.
2. Stratified sampling is the preferred method for skewed data.
3. Regularities in the data are a problem for random sampling methods.
4. Sampling is a form of feature reduction.
- The goal of a good sampling method is to achieve a representative sample.
- Stratified sampling is the preferred method for skewed data.
Was ist stratified sampling?
Stratified sampling is a method of sampling from a population by dividing it into subpopulations, or strata, and then selecting samples from each stratum independently.
Which of the following statements on quality measurements are correct?
1. Precision and recall measure opposing goals for model quality.
2. Specificity is also known as the True Positive recognition rate.
3. The accuracy measures the percentage of correctly classified evaluation samples.
4. Accuracy can be used as performance measurement for classification problems when class distribution is balanced, whereas F1-Score can be used when there are imbalanced classes in the data.
1, 3 und 4
