Mentatcurated
▸ Concept also: lossless compression, lossy compression, entropy coding, source coding

Data compression

Encoding information in fewer bits by exploiting patterns and redundancy in the source data.

In a nutshell

Every file contains patterns — repeated sequences, skewed symbol frequencies, structural regularities. Compression algorithms find those patterns and replace them with shorter representations, then reverse the process on decompression. Lossless compression recovers the original exactly; lossy compression discards detail the receiver won't miss. The hard limit is set by Shannon's entropy: no scheme can compress below the true information content of the source. Extreme compression ratios are therefore evidence of extreme redundancy — or of a model that has learned the source's structure well enough to predict it.

Where it came from

Year1948
SourceClaude Shannon, 'A Mathematical Theory of Communication', Bell System Technical Journal
Why it matteredEstablished the theoretical ceiling; practical algorithms such as Huffman coding (1952) and LZ77 (1977) followed.

How this connects

Tap a node to open it