Fit

Question 1

Model Fitness

What is bias in a model?

Answer 1

Difference between predicted and actual value

Question 2

Model Fitness

What is variance in a model?

Answer 2

How much the regression changes with different sets of training data

Question 3

Model Fitness

What is a model’s “fit”?

Answer 3

Ability to perform well (has to do with training effectiveness)

Question 4

Model Fitness

What are the two extremes and middle of the “fit” range?

Answer 4

Under-fit, balanced, over-fit

Question 5

Underfit Model

What is under-fit model?

Answer 5

The regression is too simple, it isn’t capturing enough data to be effective.

Question 6

Underfit Model

Bias and variance for an under-fit model?

Answer 6

Bias is high and the variance is low

Question 7

Underfit Model

What is a 2-D visual for an under-fit regression?

Answer 7

Straight line through all observations – too general, bad predictor

Question 8

Underfit Model

How does an under-fit model do predicting from training data?

Answer 8

Poor – the model isn’t specific enough

Question 9

Underfit Model

How does an under-fit model do predicting from new data it hasn’t seen before?

Answer 9

Also poor – the model isn’t specific enough

Question 10

Overfit Model

What is over-fit model?

Answer 10

The regression is too complex, it tightly fits the training data too much

Question 11

Overfit Model

Bias and variance for an over-fit model?

Answer 11

Bias is low and the variance is high

Question 12

Overfit Model

What is a 2-D visual for an over-fit regression?

Answer 12

Regression line directly through every training point – it’s fit to every outlier!

Question 13

Overfit Model

How does an over-fit model do predicting from training data?

Answer 13

Perfect

Question 14

Overfit Model

How does an over-fit model do predicting from new data it hasn’t seen before?

Answer 14

Terrible – it has too much noise built-in

Question 15

Balanced Model

What is a balanced model?

Answer 15

The regression captures trends, ignores outliers

Question 16

Balanced Model

Bias and variance for a balanced model?

Answer 16

Bias is low and the variance is low

Question 17

Balanced Model

What is a 2-D visual for a balanced regression?

Answer 17

Non-linear curve through the largest clusters

Question 18

Balanced Model

How does a balanced model do predicting from training data?

Answer 18

Good

Question 19

Balanced Model

How does a balanced model do predicting from new data it hasn’t seen before?

Answer 19

Good