CPRS 47: Some other Precision Medicine

Question 1

What test is this
Let the interrogator distinguish whether A or B is human or AI. If can’t distinguish, the machine is said to have intelligence

Answer 1

Turing Test

Question 2

‘Relationships to be extracted from the data without programming’
Name a specific terminology

Answer 2

Machine Learning

Question 3

What is the difference between expert learning and supervised learning

Expert: Requires explicit programming using ___________

Supervised: Machine identifying a _______ to describe the ___________

Answer 3

Expert: Requires explicit programming using pre-defined rules
Supervised: Machine identifying a function to describe the relationship

Question 4

What is the difference between supervised learning and unsupervised learning?

Supervised: Machine will identify a _______ to describe the relationship

Unsupervised: Machine identify patterns from _______ data, using strategies like _________

Answer 4

What is the difference between supervised learning and unsupervised learning?

Supervised: Machine will identify a function to describe the relationship

Unsupervised: Machine identify patterns from unlabelled data, using strategies like clustering

Question 5

Converting raw data into numerical features

What is this process called?

Answer 5

Feature extraction

Question 6

Reducing the number of input variables to have smallest number of discriminative features
What is this process called

Answer 6

Feature selection

Question 7

What is important during cross-validation

Say, when we divide data into 5 equal sized subsets, we use 4 sets to _______ the classifier, and remaining 1 set to _______ the trained classifier.

Answer 7

What is important during cross-validation

Say, when we divide data into 5 equal sized subsets, we use 4 sets to train the classifier, and remaining 1 set to evaluate the trained classifier.

Question 8

What will happen if feature selection occurs prior to cross validation?

What shall we do instead?

Answer 8

What will happen if feature selection occurs prior to cross validation? Overestimation of accuracy

(Due to unseen class label in cross-validation is used during the feature selection process already)

Perform cross-validation prior to feature selection