How Is Compression Prediction?

The article discusses the concept of compression being equivalent to prediction, as outlined by Shannon's connection of probability to optimal code length and the relationship between learning and compression. The author agrees with this correspondence, but emphasizes that it only applies once certain conditions are met, such as the encoder and decoder agreeing on the representation of the data, the availability of the probability model, and the decoder's ability to utilize the representation.

The author also highlights that compression can be understood before introducing a sequential model, and that a finite family of admissible objects provides a counting lower bound without identifying the next symbol. In summary, the article stresses that while compression and prediction are mathematically equivalent, there are specific factors that must be considered to fully apply this equivalence in practice.

Brief written by urgent.news from Lobsters's own syndicated text. Machine-written — may contain errors; check the original before relying on it.