Hands-On Machine Learning with Scikit-Learn & TensorFlow Exercise Q&A Chapter01

Q1. How would you define Machine Learning?

A1: The science to programming computers to learn from data.

 

Q2. Can you name four types of problems where it shines?

A2: Predict the weather; segment different people; speech recognization; image recognization.

 

Q3. What is a labeled training set?

A3: The labeled training set contains the desired solution for each instance.

 

Q4. What are the two most common supervised tasks?

A4: Regression; Classification.

 

Q5. Can you name four common unsupervised tasks?

A5: Dimensionality reduction; anomaly detection; clustering; association rule learning.

 

Q6. What type of Machine Learning algorithm would you use to allow a robot to walk in various unknown terrains?

A6: Reinforcement learning.

 

Q7. What type of algorithm would you use to segment your customers into multiple groups?

A7: Clustering.

 

Q8. Would you frame the problem of spam detection as a supervised learning problem or an unsupervised learning problem?

A8: Supervised learning.

 

Q9. What is an online learning system?

A9: Train the system incrementally by feeding data instances sequentially.

 

Q10. What is out-of-core learning?

A10: Train systems on huge datasets that cannot fit in one machine's main memory. So out-of-core learning algorithm loads part of the data, runs a training step on that data. and repeats the process until it has run on all of the data.

 

Q11. What type of learning algorithm relies on a similarity measure to make predictions?

A11: Instance-based learning.

 

Q12. What is the difference between a model parameter and a learning algorithm's hyperparameter?

A12: A model parameter can help the model to make predictions, while a hyperparameter is a parameter of the algorithm itself, not the model.

 

Q13. What do model-based learning algorithms search for? What is the most common strategy they use to secceed? How do they make predictions?

A13: It search for an optimal value to generalize better; minimize the cost function; feed the new instance into the prediction function.

 

Q14. Can you name four of the main challenges in Machine Learning?

A14: Poor-quality data; insufficient quantity of training data; nonrepresentative training data; irrelevant features.

 

Q15. If your model performs great on the training data but generalizes poorly to new intances, what is happening? Can you name three possible solutions?

A15: It's overfitting;3 possible solutions:gather more training data; reduce the model's complexity; select fewer parameters.

 

Q16. What is a test set and why would you want to use it?

A16: Test set is for examination of your model; we use it because it can evaluate the model.

 

Q17. What is the purpose of a validation set?

A17: Make sure the model can perform well on new data.

 

Q18. What can go wrong if you tune hyperparameters using the test set?

A18: Maybe overfitting the test set and the prediction is inaccurate.

 

Q19. What is cross-validation and why would you prefer it to a validation set?

A19: Cross-validation splits the training set into complementary subsets, each model is trained by the different combinations of these subsets; it can avoid wasting too much training data.