t-SNE for visualization in ML Python - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Understanding t-SNE's Purpose

What is the main goal of using t-SNE in data analysis?

A To perform linear regression on high-dimensional data

B To increase the number of features in the dataset for better model training

C To reduce high-dimensional data to a lower dimension for visualization while preserving local structure

D To cluster data points into predefined groups based on labels

💡 Hint

Think about what t-SNE does to data dimensions and what it tries to keep intact.

t-SNE reduces data dimensions to 2 or 3 for visualization, focusing on preserving local neighbor relationships.

❓ Predict Output

intermediate

t-SNE Output Shape

Given the following Python code using sklearn's t-SNE, what is the shape of tsne_result?

ML Python

from sklearn.manifold import TSNE
import numpy as np

X = np.random.rand(100, 50)  # 100 samples, 50 features
model = TSNE(n_components=2, random_state=42)
tsne_result = model.fit_transform(X)
print(tsne_result.shape)

A (100, 2)

B (50, 2)

C (2, 100)

D (100, 50)

💡 Hint

Remember, t-SNE reduces features but keeps the number of samples.

t-SNE transforms 100 samples from 50 features to 2 features, so the output shape is (100, 2).

❓ Hyperparameter

advanced

Effect of Perplexity in t-SNE

Which statement best describes the effect of increasing the perplexity parameter in t-SNE?

A Higher perplexity reduces the number of iterations needed for convergence

B Higher perplexity decreases the number of output dimensions

C Higher perplexity increases the learning rate automatically

D Higher perplexity considers more neighbors, leading to a more global view of the data structure

💡 Hint

Think about how perplexity relates to neighborhood size in t-SNE.

Perplexity controls the balance between local and global aspects by setting how many neighbors each point considers; higher values mean more neighbors.

❓ Metrics

advanced

Evaluating t-SNE Visualization Quality

Which metric is commonly used to evaluate how well t-SNE preserves local structure in the reduced space?

A K-nearest neighbor preservation score

B Mean squared error between original and reduced data

C Silhouette score of clusters in original high-dimensional space

D Accuracy of a classification model trained on reduced data

💡 Hint

Consider metrics that measure neighborhood consistency after reduction.

K-nearest neighbor preservation score measures how many neighbors remain the same after dimensionality reduction, reflecting local structure preservation.

🔧 Debug

expert

Identifying t-SNE Runtime Error

What error will this code raise and why?

from sklearn.manifold import TSNE
import numpy as np

X = np.random.rand(10, 5)
model = TSNE(n_components=3, random_state=0)
result = model.fit_transform(X)
print(result.shape)

A ValueError because n_components cannot be greater than 2 for t-SNE

B No error, output shape will be (10, 3)

C TypeError because input data X is not a pandas DataFrame

D RuntimeWarning due to random_state not being set

💡 Hint

Check the allowed output dimensions for t-SNE in sklearn.

No error: sklearn's TSNE supports n_components=1, 2, 3 (and higher with warnings for >3). The output shape is (10, 3).