Early stopping
Technique
In machine learning, early stopping is a form of regularization used to avoid overfitting when training a model with an iterative method, such as gradient descent. Such methods update the model to make it better fit the training data with each iteration. Up to a point, this improves the model's performance on data outside of the training set. Past that point, however, improving the model's fit to the training data comes at the expense of increased generalization error. Early stopping rules provide guidance as to how many iterations can be run before the learner begins to over-fit. Early stopping rules have been employed in many different machine learning methods, with varying amounts of theoretical foundation.
Related work
- Handwritten digit OCR Oct 2024
The notebooks compare training set-ups on the same task: a CNN trained for 50 and for 200 epochs; the same with early stopping, which ends training when validation stops improving; transfer learning from VGG16, a network pre-trained on ImageNet, including a variant with a higher learning rate; a notebook comparing the results.
As data
Everything on this page is read from triples in the knowledge graph. These are the 3 it states about Early stopping, and 1 more its ontology implies; the console runs SPARQL over the whole graph, here in your browser.
@prefix id: <https://ask.chaipat.cc/id/> . @prefix rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#> . @prefix cj: <https://ask.chaipat.cc/ns#> . @prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#> . @prefix schema: <https://schema.org/> . id:EarlyStopping a cj:Technique ; rdfs:label "Early stopping" ; schema:sameAs <https://en.wikipedia.org/wiki/Early_stopping> .