Skip to main content
clipping

clipping

3 7.0
The act of limiting large changes to a computer model during training
  • noun
  • /ˈklɪpɪŋ/
  • Specialized
translation icon : recorte de gradiente
  • There are a large number of distortion products generated by the clipping, with major peaks found in the vicinity of the third and fifth harmonics of the excitation sinusoids.
  • clipping threshold

Examples

  • During model training, clipping is essential to prevent the gradients from exploding and causing instability.

  • At higher stimulus intensities, the amplifier clipping becomes increasingly more prominent, and the output saturation is reflected in the indicated high compression ratios.

    Academic text (1990)
  • The power-amplifier behavior is represented as a symmetric clipping stage followed by a fixed gain.

    Academic text (1990)
  • By properly using and reading the resultant histogram, the photographer can ensure that the image is properly exposed with no clipping.

    Academic text (2008)
  • The amplifier distortion is simulated as symmetric clipping of either the voltage or the current, depending on the type of source selected.

    Academic text (1990)
  • The response to the same tone burst was then repeated for the same hearing aid, but with amplifier clipping at an input-referred peak level of 85 SPL.

    Academic text (1990)
  • He implemented clipping to control the sizes of the gradients during the neural network's training process.

  • The clipping technique helped maintain stable updates in the machine learning algorithm, avoiding drastic changes.

Compounds

A small piece cut from a newspaper that people keep to remember something

press clipping
  • Literature
3 3.2

A short article cut from a newspaper or magazine kept for reference

How Distorted

clipping
  • Specialized
3 7.0
amplitude distortion
  • Specialized
7.1
distortion
  • Specialized
53 6.7

Surface Forms

clipping singular
clippings plural

Morphology

clipping = clip (semi-transparent) = clip + ing

The technical ML sense is a metaphorical extension of 'clip' meaning 'limit' or 'truncate', so the formation is regular but the specialized meaning may not be obvious to general learners.

Etymology

Clipping comes from clip 'to cut off' or 'trim', and in machine learning it means cutting large numbers down so they do not grow too big during training. So, clipping keeps values within a safe size.