ML

Question 1

What training is used for datasets without labels?

Answer 1

Unsupervised training

Question 2

What training is used for datasets with labels?

Answer 2

Supervised

Question 3

What training is used for smaller datasets with labels and bigger ones wo labels?

Answer 3

Semi-supervised

Question 4

What training is used when positive and negative feedbacks are given?

Answer 4

Reinforcement

Question 5

What training is used based on previous knowledge?

Answer 5

Trasnfer

Question 6

What algorithm predicts the numerical value of a variable based on independent variables?

Answer 6

Linear regression

Question 7

What algorithm predicts a binary outcome of independent variables?

Answer 7

Logistic regression (binary)

Question 8

What algorithm predicts multiple outcomes of independent variables?

Answer 8

Logistic regression (multinomial)

Question 9

What algorithm predicts outcome with order of independent variables?

Answer 9

Logistic regression (ordinal)

Question 10

What algorithm predicts results on data with curvilineal patterns?

Answer 10

Polinomial regression

Question 11

What algorithms try to decrease overfitting in linear regression?

Answer 11

Ridge and Lasso regressions

Question 12

What is the name of how much changing training sets change the model?

Answer 12

Variance

Question 13

What is the name of the phenomenon that measures how much the training set is different from real world scenarios?

Answer 13

Bias, vies

Question 14

Simplistic models tend to have what kind of bias variance trade-off?

Answer 14

High Bias and low variance

Question 15

Complex models tend to have what kind of bias variance trade-off?

Answer 15

Low bias and high variance

Question 16

Which bias-variance trade off results in Underfitting?

Answer 16

High Bias and low variance

Question 17

Which bias-variance trade off results in Overfitting?

Answer 17

Low Bias and high variance

Question 18

What is the name of the specific table layout that allows visualization of the performance of an algorithm in ML?

Answer 18

Confusion matrix

Question 19

In an error matrix, what is the name of a prediction that is 1 when the actual classification is 0?

Answer 19

False positive

Question 20

In an error matrix, what is the name of a prediction that is 0 when the actual classification is 1?

Answer 20

False negative

Question 21

Which dimension reduction algorithm is used to simplify the data model while keeping the variance?

Answer 21

Principal component analysis PCA

Question 22

What kind of supervised algorithm is used for classification and regression of multi-dimensional data models that cannot be separated linearly?

Answer 22

Support Vector Machines SVM

Question 23

Which algorithm uses multiple decision trees trained on different subsets of data and then use a combination of their results?

Answer 23

Random forests

Question 24

Which clustering algorithm groups elements in levels so that it iterates grouping or dividing clusters in a tree structure?

Answer 24

Hierarchical clustering

Question 25

Which algorithm uses a majority of the closest nodes to define a classification?

Answer 25

K nearest neighbors

Question 26

How knn makes regression?

Answer 26

It uses commonly an average of the distance of the nearest neighbors

Question 27

What heapens if we chose a low K value in knn?

Answer 27

Low K results in noise sensitivity, which is high variance and low bias, thus overfitting

Question 28

What heapens if we chose a high K value in knn?

Answer 28

High K values result in oversimplification, so high bias and low variance, underfitting

Question 29

Which algorithm uses iterative centroids that get closer to clusters in a unsuperviserd learning?

Answer 29

K-means