Open Models Almanac
Methodsdata curationtokenization

single mention · data curation

Discrete cosine transform (DCT) action compression

Discrete cosine transform (DCT) action compression is a frequency-space transform that represents a continuous signal as a sum of cosine elements of various frequencies.1 It compresses signals by omitting frequency components with low weights, so most information fits in a few coefficients, making it a simple and fast analytical alternative to learned compression.2

Trade-off The scaling coefficient trades off between lossiness and compression rate of the tokenization.3

How this definition was written

Summarised by glm-5.3-flash from 1 source; every sentence is checked against the quote it cites.

  1. DCT is a frequency-space transform that represents a continuous signal as a sum of cosine elements of various frequencies — FAST: Efficient Action Tokenization for Vision-Language-Action Models Thanks: ∗: Core contributors Correspondence to:…
  2. Signals can be compressed by omitting frequency components with low weights. Compared to learned compression approaches based on vector quantization, DCT-based compression is an analytical approach, thus extremely simple and fast — FAST: Efficient Action Tokenization for Vision-Language-Action Models Thanks: ∗: Core contributors Correspondence to:…
  3. the scaling coefficient is a hyperparameter that trades off between lossiness and compression rate of the tokenization operation — FAST: Efficient Action Tokenization for Vision-Language-Action Models Thanks: ∗: Core contributors Correspondence to:…

Also called discrete cosine transform (DCT).

Filed under data curation › tokenization