Learning curves in ML Python - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Understanding the shape of learning curves

Which of the following best describes the typical shape of a learning curve for a well-trained machine learning model as the training data size increases?

A Both training and validation errors decrease and then plateau as data size grows.

B The training error increases and the validation error decreases steadily.

C Training error remains constant while validation error fluctuates randomly.

D Training error increases sharply while validation error remains zero.

❓ Predict Output

intermediate

Output of learning curve plot code

What will be the output of the following Python code snippet using scikit-learn's learning_curve function?

ML Python

from sklearn.datasets import load_iris
from sklearn.svm import SVC
from sklearn.model_selection import learning_curve
import numpy as np

X, y = load_iris(return_X_y=True)
train_sizes, train_scores, test_scores = learning_curve(SVC(kernel='linear'), X, y, train_sizes=np.linspace(0.1, 0.5, 5), cv=3)
print(train_sizes)
print(np.round(train_scores.mean(axis=1), 2))
print(np.round(test_scores.mean(axis=1), 2))

A

[15 30 45 60 75]
[0.8 0.85 0.9 0.95 1. ]
[0.7 0.75 0.8 0.85 0.9]

B

[15 30 45 60 75]
[0.5 0.5 0.5 0.5 0.5]
[0.5 0.5 0.5 0.5 0.5]

C

[10 20 30 40 50]
[1. 1. 1. 1. 1.]
[1. 1. 1. 1. 1.]

D

[15 30 45 60 75]
[1.   0.98 0.97 0.96 0.95]
[0.95 0.94 0.93 0.92 0.91]

❓ Hyperparameter

advanced

Effect of model complexity on learning curves

How does increasing model complexity typically affect the shape of training and validation learning curves?

A Both training and validation errors increase with model complexity.

B Training error decreases but validation error first decreases then increases with complexity.

C Training error increases and validation error decreases as complexity grows.

D Training error decreases and validation error decreases steadily with complexity.

❓ Metrics

advanced

Interpreting learning curve metrics

If a learning curve shows a large gap between training and validation errors that does not decrease with more data, what does this indicate?

A The model is perfectly fitted with no issues.

B The model is underfitting and needs more features.

C The model is overfitting and more training data might help.

D The data is too noisy and cannot be learned.

🔧 Debug

expert

Diagnosing incorrect learning curve results

A user runs learning_curve on a classification task but notices training scores are always lower than validation scores. What is the most likely cause?

A The scoring metric is incorrectly set to a metric that favors validation data.

B The model is too simple and underfits the training data.

C The training data is smaller than validation data causing unstable training scores.

D The cross-validation splits are not stratified, causing unbalanced classes in training.