How Compression Works | Further Reading Index

Continue from How Compression Works. Choose a topic below to explore the specialist studies, examples and limits in more detail.

The full cost of a smaller representation

• Borrowed Time — Why Saving Space Can Make the Future Do More Work
• The Model Is Part of the Message — When a Smaller Description Needs a Bigger Theory
• Locality — Why Making Everything Smaller Can Make One Small Thing Harder to Reach
• Reconstruction — The Small File Is Only Half the Machine

Boundaries, shared knowledge and representation

• Boundaries — Why Where You Cut Information Changes How Small It Can Become
• Canonical Form — Why the Same Thing Written Two Ways Can Hide a Duplicate
• Side Information — How Knowing Something Else Can Make This Message Smaller
• Amortisation — When a Huge Dictionary Becomes Cheap Because Everyone Reuses It

Families of files and reusable structure

• The Corpus — Why One File Can Become Smaller When It Belongs to a Family
• Deduplication — When Many Files Can Point to One Shared Chunk
• Delta Encoding — Why a New Version Can Be Stored as “What Changed”
• Shared Dictionaries — How One Learned Vocabulary Can Compress a Whole Family

Memory, model choice and changing data

• Memory — How Much of the Past Must a Compressor Remember?
• Regret — How Much Do We Pay for Choosing the Wrong Model Too Early?
• Forgetting — Why Keeping Every Pattern Forever Can Make Compression Worse
• Corpus Drift — When the Family Changes and Yesterday’s Model Stops Fitting

Return to How Compression Works or explore the Information & Representation Hub.