Picking the bias the data likes best

Computing · Machine Learning · ages 18-20

Name ______________________ Date ____________

Which expression is the likelihood for seven heads in ten flips?
- 7 plus 3
- p to the 7 times (1 minus p) to the 3
- 10 divided by 7
The likelihood is a function of what?
- The parameter, with data fixed
- The data, with bias fixed
- The weather on flip day
Fitting by likelihood means picking the setting that maximizes the score.
Circle one: True False
Why check the answer against seven out of ten?
- Gut feeling replaces all math
- Gut checks catch flipped roles
- Exams award marks for hunches
Seven heads in ten flips. Type the maximizing bias.
Answer: ______________
Where does the least squares line come from?
- From coin weight alone
- From sorting residuals by color
- From maximizing a likelihood
A higher likelihood means the bias is certainly true.
Circle one: True False
A student fixes p = 0.5 and scores many datasets. What is backwards?
- The data should be fixed while the bias moves
- Halves are never allowed
- Datasets must be scored one by one aloud

Answer key

For grown-ups. Fold this page away before handing over the rest.

p to the 7 times (1 minus p) to the 3 · Seven heads supply p seven times, three tails the rest.
The parameter, with data fixed · Bias moves while flips stay put.
True · Top score wins the fit.
Gut checks catch flipped roles · Intuition confirms the peak sits in a sane place.
0.7 · Seven out of ten peaks at seven tenths.
From maximizing a likelihood · It is one child of the likelihood rule.
False · It means best scoring, not proven true.
The data should be fixed while the bias moves · Likelihood moves the parameter, not the data.