PCA Intuition
Principal component analysis (PCA) is a powerful technique that identifies new axes - called principal components - which are directions in your data that capture the most variance.
PCA keeps the directions where your data varies the most, as these capture the key patterns and structure.
Think of PCA like shining a flashlight on a 3D object and examining the shadow on a wall. The angle of the light changes the shadow's detail. PCA finds the best angle so the shadow, or projection, reveals the most about the object's shape. Similarly, PCA projects your data onto new axes to preserve as much variation as possible.
12345678910111213141516171819202122232425262728293031323334import numpy as np import matplotlib.pyplot as plt # Generate a simple 2D dataset np.random.seed(0) mean = [0, 0] cov = [[3, 2], [2, 2]] # Covariance matrix X = np.random.multivariate_normal(mean, cov, 200) # Compute the mean of the data mean_vector = np.mean(X, axis=0) # Compute the covariance matrix and its eigenvectors cov_matrix = np.cov(X.T) eigenvalues, eigenvectors = np.linalg.eig(cov_matrix) # First principal component (direction of maximum variance) pc1 = eigenvectors[:, np.argmax(eigenvalues)] # Plot the data plt.figure(figsize=(8,6)) plt.scatter(X[:,0], X[:,1], alpha=0.3, label="Data points") plt.quiver( mean_vector[0], mean_vector[1], pc1[0], pc1[1], angles='xy', scale_units='xy', scale=1.5, color='red', width=0.01, label="First principal component" ) plt.xlabel("Feature 1") plt.ylabel("Feature 2") plt.title("Direction of Maximum Variance (First Principal Component)") plt.legend() plt.axis("equal") plt.show()
By identifying the directions where your data varies the most, PCA allows you to reduce dimensions while preserving the most important information. Focusing on these directions of maximum variance ensures that the structure and patterns in your dataset remain clear. This understanding prepares you to explore the mathematical foundation of PCA in upcoming sections.
Thanks for your feedback!
Ask AI
Ask AI
Ask anything or try one of the suggested questions to begin our chat
Awesome!
Completion rate improved to 8.33PCA Intuition
Swipe to show menu
Principal component analysis (PCA) is a powerful technique that identifies new axes - called principal components - which are directions in your data that capture the most variance.
PCA keeps the directions where your data varies the most, as these capture the key patterns and structure.
Think of PCA like shining a flashlight on a 3D object and examining the shadow on a wall. The angle of the light changes the shadow's detail. PCA finds the best angle so the shadow, or projection, reveals the most about the object's shape. Similarly, PCA projects your data onto new axes to preserve as much variation as possible.
12345678910111213141516171819202122232425262728293031323334import numpy as np import matplotlib.pyplot as plt # Generate a simple 2D dataset np.random.seed(0) mean = [0, 0] cov = [[3, 2], [2, 2]] # Covariance matrix X = np.random.multivariate_normal(mean, cov, 200) # Compute the mean of the data mean_vector = np.mean(X, axis=0) # Compute the covariance matrix and its eigenvectors cov_matrix = np.cov(X.T) eigenvalues, eigenvectors = np.linalg.eig(cov_matrix) # First principal component (direction of maximum variance) pc1 = eigenvectors[:, np.argmax(eigenvalues)] # Plot the data plt.figure(figsize=(8,6)) plt.scatter(X[:,0], X[:,1], alpha=0.3, label="Data points") plt.quiver( mean_vector[0], mean_vector[1], pc1[0], pc1[1], angles='xy', scale_units='xy', scale=1.5, color='red', width=0.01, label="First principal component" ) plt.xlabel("Feature 1") plt.ylabel("Feature 2") plt.title("Direction of Maximum Variance (First Principal Component)") plt.legend() plt.axis("equal") plt.show()
By identifying the directions where your data varies the most, PCA allows you to reduce dimensions while preserving the most important information. Focusing on these directions of maximum variance ensures that the structure and patterns in your dataset remain clear. This understanding prepares you to explore the mathematical foundation of PCA in upcoming sections.
Thanks for your feedback!
