Quiz on statistical learning framework, empirical minimization framework and PAC learning. Machine learning is a formal learning model.   Statistical learning is a subfield of machine learning that uses statistical methods to identify patterns and relationships in data. Statistical learning is the basis for machine learning algorithms.  The Empirical Risk Minimization (ERM) principle is a learning paradigm which consists in selecting the model with minimal average error over the training set. This so-called training error can be seen as an estimate of the risk (due to the law of large... Show more

Quiz on statistical learning framework, empirical minimization framework and PAC learning.

Machine learning is a formal learning model.
 
Statistical learning is a subfield of machine learning that uses statistical methods to identify patterns and relationships in data. Statistical learning is the basis for machine learning algorithms. 

The Empirical Risk Minimization (ERM) principle is a learning paradigm which consists in selecting the model with minimal average error over the training set. This so-called training error can be seen as an estimate of the risk (due to the law of large numbers), hence the alternative name of empirical risk.

Probably Approximately Correct (PAC) learning is a mathematical framework for analyzing machine learning. It was proposed in 1984 by Leslie Valiant. 

In PAC learning, the learner receives samples and selects a generalization function from a class of possible functions. The goal is to find a hypothesis, or model, that approximates the target concept as accurately as possible.

Show less

Machine Learning 101 Practice Test: Formal Learning Model

Time left 00:00

25 Questions

1. When was PAC learning invented?

1984188418741974

2. What is the significance of epsilon in PAC learning?

Probability of approximation < epsilonMinimum error > epsilonMaximum error < epsilonProbability of approximation = delta – epsilon

3. What is the relation between Empirical Risk Minimization and Training Error?

It depends on the datasetERM is not concerned with training errorERM tries to maximize training errorERM tries to minimize training error

4. What is the significance of delta in PAC learning?

Error < deltaProbability of approximation < deltaProbability = 1 – deltaProbability of approximation = delta – epsilon

5. What is assumed while using empirical risk minimization with inductive bias?

No assumptions are madeReduction of overfitting may lead to underfittingThe learner has some knowledge about labeling functionThe learner has some prior knowledge about training data

6. What is not accessible to the learner?

Training SetLabeling FunctionLabel SetDomain Set

7. In PAC learning, sample complexity grows as the logarithm of the number of hypothesizes.

TrueFalseDon't knowMaybe

8. What are the possible values of A, B, and C in the following diagram?

A – Training Set, B – Domain Set, C – Cross-Validation SetA – Test Set, B – Domain Set, C – Training SetA – Training Set, B – Test Set, C – Cross-Validation SetA – Training Set, B – Test Set, C – Domain Set

9. Which is the more desirable way to reduce overfitting?

Giving an upper bound to the size of the training setOverfitting cannot be reducedGiving an upper bound to the accuracy obtained on the training setMaking the test set larger than the training set

10. One of the goals of PAC learning is to give __________

error of classifiermaximum accuracycomputational complexitycross-validation complexity

11. The labeling function is known to the learner in the beginning.

FalseDon't knowMaybeTrue

12. The true error is available to the learner.

FalseTrueDon't knowMaybe

13. What is one of the drawbacks of Empirical Risk Minimization?

OverfittingNo drawbacksUnderfittingBoth Overfitting and Underfitting

14. The error available to the learner is ______

training errortrue errortesting errorerror of the classifier

15. The full form of PAC is ______

Partly Approximation ComputationProbability Approximation CurvePartly Approximately CorrectProbably Approximately Correct

16. What happens due to overfitting?

Hypothesis works well on training data and works well on test dataHypothesis works poorly on training data but works well on test dataHypothesis works poorly on training data and works poorly on test dataHypothesis works well on training data but works poorly on test data

17. The error of classifier is measured with respect to _________

variance of data instancesprobability distributionlabeling functionprobability distribution and labeling function

18. Number of hypothesizes |H| = 973, probability = 95%, error < 0.1. Find minimum number of training examples, m, required.

99.89998.898

19. The hypothesis space H for inductive bias is a finite class.

MaybeTrueFalseDon't know

20. Who introduced the concept of PAC learning?

Reverend Thomas BayesFrancis GaltonLeslie ValiantJ.Ross Quinlan

21. The set which represents the different instances of the target variable is known as ______

training settest setdomain setlabel set

22. A learner can be deemed consistent if it produces a hypothesis that perfectly fits the __________

training datatest dataoverall datasetcross-validation data

23. What is the learner’s output also called?

Predictor, or Trainer, or ClassifierPredictor, or Hypothesis, or ClassifierPredictor, or Hypothesis, or TrainerTrainer, or Hypothesis, or Classifier

24. What can be explained by PAC learning?

UnderfittingSample ComplexityLabel FunctionOverfitting

25. Delta is the __________ parameter of the prediction.

trainingaccuracycomputingconfidence