🚪 To quit early 🌄 To leave society 🤖 To deactivate units

to drop out

2 5.9

To stop using some parts of a machine learning model at random during training so it learns better

phrasal verb

Specialized

: desactivar

The long-term memory stores the information for retrieval if it already has associations in the brain from prior learning that provide anchors, or the data drops out.

And So what you do is there's various strategies you can do drop out where you, you to periodically turn off half of the nodes in a network.

Examples

The short-term memory, within 20-30 seconds, encodes the data and sends it to the long-term memory system, or it drops out.

Academic text (1997)
It randomly chooses units to drop out and removes them from the layer temporarily.

Academic text (2018)
So you see half of the neurons drop out and then I'm going to have to chip Google itself.
During training, the model can randomly drop out some neurons to help improve its performance.
To avoid overfitting, we can drop out a percentage of the neurons in each training iteration.
In this layer, certain neurons will drop out randomly, which helps the model generalize better.

Antonyms

to activate

715 3.6

To make a machine, system, or account start working

Surface Forms

drop out infinitive

dropped out past

dropped out past participle

dropping out present participle

drops out third person singular

Morphology

drop + out

The ML sense is a domain-specific application of the basic phrasal-verb sense 'drop out' = 'remove or exclude'. A learner who knows 'drop' (to remove/stop including) and 'out' (excluded) can compositionally infer that certain units are being excluded/deactivated, so the meaning is transparently derived from the constituents.

Etymology

Drop out in machine learning comes from the idea of letting some parts drop out of a lesson so the rest must learn without them, so it means 'to turn off some neurons during training' to help the model learn better.