Qwissit Artificial Intelligence

ISTQB CT-AI

Machine Learning and Data

60 questions

Introduction to machine learning

  1. Into which three forms can ML algorithms be categorized? Easy
  2. A model is to classify emails as spam or not spam. What kind of problem is this? Easy
  3. A model is to estimate a person's age from lifestyle data. What kind of problem is this? Easy
  4. How does the term "regression" in the ML context differ from its meaning in classic software testing? Easy
  5. What feedback does an intelligent agent receive for its actions in reinforcement learning? Easy
  6. What happens in the first activity of the ML workflow, "understand the objectives"? Easy
  7. Which parameters are called model hyperparameters? Easy
  8. Which dataset is used to assess the model in the activity "evaluate the model"? Easy
  9. Which three activities together make up "model generation"? Easy
  10. How is the ML workflow applied in practice? Easy
  11. What is meant by fine-tuning? Easy
  12. What does the abbreviation RAG stand for? Easy
  13. What happens to the pretrained model with RAG? Easy

Data for machine learning

  1. How much effort does data preparation take in the ML workflow compared with other steps? Easy
  2. Which three main activities does data preparation comprise? Easy
  3. Which techniques are typically used to fill in missing values during data cleaning (imputation)? Easy
  4. What is the goal of exploratory data analysis (EDA)? Easy
  5. What is the validation dataset used for? Easy
  6. What is the test dataset also called? Easy
  7. Which values of k are common in k-fold cross-validation? Easy

Introduction to machine learning

  1. An online retailer wants to derive from purchase data which products are often bought together in order to make recommendations. Which form of ML fits? Medium
  2. Which challenges does reinforcement learning involve? Medium
  3. What does the choice of ML form (supervised, unsupervised, reinforcement) depend on? Medium
  4. Which parameters are called algorithm hyperparameters? Medium
  5. What does the activity "prepare & test data" in the ML workflow comprise? Medium
  6. How is the best model usually found in practice? Medium
  7. What needs to be done if the model performs significantly worse on the independent test dataset than during evaluation? Medium
  8. Who typically performs the testing of the model in the ML workflow? Medium
  9. What happens in the activity "deploy the model"? Medium
  10. What does the activity "monitor & tune the model" involve? Medium
  11. What is not included in the ML workflow shown, but usually has to be done additionally? Medium
  12. To which parts of a neural network can fine-tuning be applied? Medium
  13. What does the success of fine-tuning mainly depend on? Medium
  14. What happens to biases and vulnerabilities of a pretrained model when it is adapted to a new task? Medium
  15. How does RAG work? Medium

Data for machine learning

  1. Which tasks belong to data acquisition? Medium
  2. Which tasks belong to data preprocessing? Medium
  3. Why should the operational data match the characteristics of the training data? Medium
  4. Which factors determine the amount of data used for training, evaluation and testing when enough data is available? Medium
  5. How does k-fold cross-validation work? Medium

Introduction to machine learning

  1. Which mappings of task and ML form are correct? Hard
  2. A tester reads in an ML project report: "Regression completed, error 3.2 years". How is this most likely to be understood? Hard
  3. A team has trained three models with random forest, SVM and a neural network and chosen the best one based on the validation data. Which step is missing before deployment? Hard
  4. After tuning, a model achieves 94% accuracy on the validation data but only 78% on the independent test dataset. Which conclusions are appropriate? Hard
  5. Which examples belong to model hyperparameters and which to algorithm hyperparameters? Hard
  6. An image processing model is to be built into vehicle software. Which statements about integration are correct? Hard
  7. A model has been running in production for months; its precision is gradually falling below the acceptance criterion. What should happen next? Hard
  8. A model has been successfully tested and is now being re-engineered for an embedded system. What is needed afterwards? Hard
  9. A classifier for apple varieties is to be adapted by fine-tuning. Which target tasks are likely to work well and which less well? Hard
  10. A company wants an LLM to answer questions about internal manuals without changing the model itself. What is recommended, and what must be considered? Hard
  11. Which statements about the relationship between fine-tuning and RAG are correct? Hard

Data for machine learning

  1. A fraud detection model was trained with data from Europe and is now used in Asia with significantly different transaction patterns. What is the consequence? Hard
  2. Which tasks belong to feature engineering? Hard
  3. Which statements about the course of data preparation are correct? Hard
  4. A team deliberately adds adversarial examples to the training data of an image classifier. How is this to be classified? Hard
  5. A team has only a few hundred records. Which steps are common in such cases? Hard
  6. The data is so scarce that no holdout test set is possible; only the average of the cross-validation is reported. How is this value to be assessed? Hard
  7. A dataset is highly imbalanced (few fraud cases). How should the folds be formed in k-fold cross-validation? Hard
  8. Which other resampling methods for scarce data exist besides k-fold cross-validation? Hard
  9. Why must the training, validation and test datasets be equivalent? Hard

Certifications & exams

Which exam comes next?

You can already practise for 19 exams, and 21 are up for a vote. See all certifications with their official information, cast your vote or suggest your exam.

See all certifications