Add Naive Bayes spam filter interactive lesson in Vietnamese

[LMK89]

Summary

Added a comprehensive interactive educational lesson on Naive Bayes classification for spam filtering, written in Vietnamese. The lesson includes theoretical explanations, practical demonstrations, and interactive quizzes.

Key Changes

• New lesson file: 2026-08-30-naive-bayes-spam-filter-201288cf.html (629 lines)

  • Complete HTML5 document with embedded CSS and JavaScript
  • Dark/light theme toggle with localStorage persistence
  • Responsive design using CSS Grid and Flexbox

• Content sections:

  • Quick warmup quiz (3 questions) to assess prior knowledge
  • Formal definition of Naive Bayes and Bayes' theorem
  • Real-world email spam classification problem with training data table
  • Four detailed state comparisons:
    • Zero-frequency problem (without Laplace smoothing)
    • Laplace smoothing solution (add-1)
    • Numerical underflow issues (direct multiplication vs log-probability)
    • Imbalanced class priors
  • Interactive demo calculator showing step-by-step probability calculations
  • Data quality guidelines (good vs bad datasets)
  • Self-grading quiz with 3 conceptual questions

• Interactive features:

  • Checkbox selection for email keywords
  • Radio buttons for smoothing mode toggle
  • Real-time calculation with animated step-by-step output
  • Random email generator for testing
  • Immediate feedback on quiz answers
  • Responsive form inputs and buttons

• Updated README.md: Added entry for the new lesson in the course index table

Implementation Details

• Uses CSS custom properties (CSS variables) for theming
• Implements smooth animations (fadeInDown, slideUp, slideDown, pulse)
• JavaScript calculations based on training data: 4 spam emails and 4 ham emails with 6 keywords
• Demonstrates both problematic (zero-frequency) and correct (Laplace smoothing) approaches
• Educational focus on practical pitfalls: smoothing, numerical stability, and class imbalance

https://claude.ai/code/session_01Dd4TgW5XR5AP1GNyEGeJ8k