15 `TableVectorizer`

15.1 Question 1

What is the proper way of setting the number of components used by the high cardinality encoder in the TableVectorizer?

A) By passing n_components as kwargs:

TableVectorizer(n_components=10)

B) By creating a new encoder with the specified parameter:

new_encoder = StringEncoder(n_components=10)
TableVectorizer(high_cardinality=new_encoder)

C) Both A) and B) are correct
D) None of the above

Tip

Answer: B)

15.2 Question 2

The TableVectorizer categorizes columns based on their characteristics. What are the types of columns that it handles?

A) Numeric, categorical, datetimes
B) Numeric, datetimes, high cardinality categorical, low cardinality categorical
C) Integer, float, string, categorical, datetimes
D) Numeric, datetimes, high cardinality categorical, low cardinality categorical, and complex objects such as structs or lists

Tip

Answer: B)

By default, the TableVectorizer converts complex objects to their string representation, before checking their cardinality. No special transformation is provided for columns that contain structs or lists.

16 `tabular_pipeline`

16.1 Question 3

What are the default estimators used by the tabular_pipeline when it is called with "regression" and "classification" respectively?

tabular_pipeline("regression")
tabular_pipeline("classification")

A) RidgeCV and ExtraTreeClassifier
B) HistGradientBoostingRegressor and LogisticRegression
C) RandomForestRegressor and RandomForestClassifier
D) HistGradientBoostingRegressor and HistGradientBoosting

Tip

Answer: D)

16.2 Question 4

How are linear models (Ridge, LogisticRegression) handled by the tabular_pipeline?

A) The TableVectorizer uses default parameters, no other steps are added to the pipeline.
B) The TableVectorizer uses default parameters, SimpleImputer and StandardScaler are added before the encoder.
C) The TableVectorizer uses a datetime encoder that periodic encodings with splines, SimpleImputer and StandardScaler are added before the encoder.
D) The TableVectorizer uses a datetime encoder that periodic encodings with splines, no other steps are added to the pipeline.

Tip

Answer: C)

--- title: "Quiz: Column-level transformations" format: html: code-tools: true --- # `TableVectorizer` ## Question 1 ::: {.callout} What is the proper way of setting the number of components used by the high cardinality encoder in the `TableVectorizer`? - [ ] A) By passing `n_components` as kwargs: ```{.python} TableVectorizer(n_components=10) ``` - [ ] B) By creating a new encoder with the specified parameter: ```{.python} new_encoder = StringEncoder(n_components=10) TableVectorizer(high_cardinality=new_encoder) ``` - [ ] C) Both A) and B) are correct - [ ] D) None of the above ::: ::: {.callout-tip collapse="true"} Answer: B) ::: ## Question 2 ::: {.callout} The `TableVectorizer` categorizes columns based on their characteristics. What are the types of columns that it handles? - [ ] A) Numeric, categorical, datetimes - [ ] B) Numeric, datetimes, high cardinality categorical, low cardinality categorical - [ ] C) Integer, float, string, categorical, datetimes - [ ] D) Numeric, datetimes, high cardinality categorical, low cardinality categorical, and complex objects such as structs or lists ::: ::: {.callout-tip collapse="true"} Answer: B) By default, the `TableVectorizer` converts complex objects to their string representation, before checking their cardinality. No special transformation is provided for columns that contain structs or lists. ::: # `tabular_pipeline` ## Question 3 ::: {.callout} What are the default estimators used by the `tabular_pipeline` when it is called with `"regression"` and `"classification"` respectively? ```{.python} tabular_pipeline("regression") tabular_pipeline("classification") ``` - [ ] A) `RidgeCV` and `ExtraTreeClassifier` - [ ] B) `HistGradientBoostingRegressor` and `LogisticRegression` - [ ] C) `RandomForestRegressor` and `RandomForestClassifier` - [ ] D) `HistGradientBoostingRegressor` and `HistGradientBoosting` ::: ::: {.callout-tip collapse="true"} Answer: D) ::: ## Question 4 ::: {.callout} How are linear models (`Ridge`, `LogisticRegression`) handled by the `tabular_pipeline`? - [ ] A) The `TableVectorizer` uses default parameters, no other steps are added to the pipeline. - [ ] B) The `TableVectorizer` uses default parameters, `SimpleImputer` and `StandardScaler` are added before the encoder. - [ ] C) The `TableVectorizer` uses a datetime encoder that periodic encodings with splines, `SimpleImputer` and `StandardScaler` are added before the encoder. - [ ] D) The `TableVectorizer` uses a datetime encoder that periodic encodings with splines, no other steps are added to the pipeline. ::: ::: {.callout-tip collapse="true"} Answer: C) :::