Custom transformers in ML Python - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

Output of a simple custom transformer

What is the output of the following code when transforming the input data?

ML Python

from sklearn.base import BaseEstimator, TransformerMixin
import numpy as np

class MultiplyByTwo(BaseEstimator, TransformerMixin):
    def fit(self, X, y=None):
        return self
    def transform(self, X):
        return X * 2

transformer = MultiplyByTwo()
input_data = np.array([[1, 2], [3, 4]])
output = transformer.transform(input_data)
print(output)

A

[[1 2]
 [3 4]]

B TypeError

C

[[0.5 1]
 [1.5 2]]

D

[[2 4]
 [6 8]]

💡 Hint

Think about what multiplying by two does to each element in the array.

The transform method multiplies each element by 2, so the output doubles each input value.

❓ Model Choice

intermediate

Choosing the right method to implement in a custom transformer

Which method must be implemented in a custom transformer to apply a transformation to the data?

A transform

B score

C predict

D fit_transform

💡 Hint

This method changes the data without learning from it.

The transform method applies the transformation to the input data. fit is for learning parameters, transform applies changes.

❓ Hyperparameter

advanced

Effect of a hyperparameter in a custom transformer

Consider a custom transformer that adds a constant value to all features. What effect does changing the constant hyperparameter have?

A It changes the amount added to each feature during transformation.

B It changes the number of features in the output.

C It changes the data type of the output features.

D It changes the order of features in the output.

💡 Hint

Think about what adding a constant means for each feature value.

The constant hyperparameter controls how much is added to each feature value during transform.

🔧 Debug

advanced

Debugging a custom transformer that fails to transform

Why does this custom transformer raise an error when calling transform? class AddOne(BaseEstimator, TransformerMixin): def fit(self, X, y=None): return self def transform(self, X): return X + 1 transformer = AddOne() transformer.transform([1, 2, 3])

A ValueError because fit was not called before transform

B AttributeError because transform method is missing

C TypeError because input is a list, not a numpy array

D No error, output is [2, 3, 4]

💡 Hint

Check the input type and what adding 1 means for a list.

Adding 1 to a list causes a TypeError because list + int is invalid. Converting input to numpy array fixes this.

🧠 Conceptual

expert

Why implement a custom transformer in a machine learning pipeline?

What is the main advantage of creating a custom transformer for a machine learning pipeline?

A To avoid using any external libraries for data processing

B To reuse specific data processing steps consistently and integrate them seamlessly in pipelines

C To replace the model training step with a simpler approach

D To automatically tune hyperparameters during training

💡 Hint

Think about how pipelines help organize repeated steps.

Custom transformers allow you to package data processing logic so it can be reused and combined with other steps in a pipeline.