Property testing algorithms for a specific property:

a) Pass inputs that have the property

b) Fail inputs that are not even close (in some predefined sense) to having the property

c) May pass or fail inputs that do not have the property, but are close to having it

d) Run in sub-linear time

e) All of the above

Property testing algorithms:

a) Are useful when an exact answer is needed

b) Are useful when when time is crucial

c) Are only useful on graphs

Name one advantage of streaming over sampling:

a) Low space usage

b) No data element is missed

c) Low running time

Which of the following situations are not amenable to coreset compression:

a) GPS travel data

b) Hospital patient logs

c) Uniformly distributed synthetic data

d) Biased uniformly distributed synthetic data

Using topic models in multi-aspect summarization increases prediction accuracy because:

a) It helps to predict the number of aspects

b) It helps to disambiguate word usage in the context of the corresponding aspects

c) It makes summaries more fluent

d) It identifies common words that can be excluded from an output summary

The method for multi-aspect summarization utilizes unlabeled data to learn:

a) The likelihood of transition between topics

b) The likelihood of sentence label given its topic and the words which it contains

c) The likelihood of a sentence given its topic

d) The likelihood of a sentence given its topic and the likelihood of transition between topics

Which of the following is most likely to provide useful information about changes in the state of a hospitalized patient?

a) Billing records

b) Pharmacy records

c) Demographic information about the patient.

Consider building a feature vector by binarizing the data in an EMR. The resulting feature vector is likely to be:

a) Poorly correlated with the patient's health status.

b) Very sparse.

c) Low dimensional.

What types of raw data was used by Khandani, Kim, and Lo (2010) in their analysis of a major U.S. bank's consumer credit card business?

a) Credit bureau data.

b) Banking and credit-card transactions data.

c) Bank balance data.

d) All of the above.

The model evaluation framework of Khandani, Kim, and Lo (2010) consists of which three distinct and non-overlapping time periods?

a) Training period, delay, and testing period.

b) Data cleaning, data compression, data archiving.

c) Analysis, estimation, and prediction.

d) Morning, noon, night.