 |
Free and accessible knowledge |
Conference papers
Efficient Model Selection for Regularized Classification by Exploiting Unlabeled Data
Abstract : Hyper-parameter tuning is a resource-intensive task when optimizing classification models. The commonly used k-fold cross validation can become intractable in large scale settings when a classifier has to learn billions of parameters. At the same time, in real-world, one often encounters multi-class classification scenarios with only a few labeled examples; model selection approaches often offer little improvement in such cases and the default values of learners are used. We propose bounds for classification on accuracy and macro measures (precision, recall, F1) that motivate efficient schemes for model selection and can benefit from the existence of unlabeled data. We demonstrate the advantages of those schemes by comparing them with k-fold cross validation and hold-out estimation in the setting of large scale classification.
Cited literature [15 references]
https://hal.archives-ouvertes.fr/hal-01237129
Contributor : Georgios Balikas Connect in order to contact the contributor
Submitted on : Wednesday, December 2, 2015 - 5:11:56 PM
Last modification on : Tuesday, January 18, 2022 - 10:30:02 AM
Long-term archiving on: : Saturday, April 29, 2017 - 5:11:58 AM
Contributor : Georgios Balikas Connect in order to contact the contributor
Submitted on : Wednesday, December 2, 2015 - 5:11:56 PM
Last modification on : Tuesday, January 18, 2022 - 10:30:02 AM
Long-term archiving on: : Saturday, April 29, 2017 - 5:11:58 AM
File
Quantification_IDA15.pdf
Files produced by the author(s)