|
ИСТИНА |
Войти в систему Регистрация |
ФНКЦ РР |
||
New Properties of the Data Distillation Method When Working with Tabular Dataстатья
- Авторы: Medvedev Dmitry, Alexander D’yakonov
- Сборник: Lecture Notes in Computer Science
- Том: 12602
- Год издания: 2021
- Издательство: Springer International Publishing AG
- Местоположение издательства: Cham, Switzerland
- Первая страница: 379
- Последняя страница: 390
- DOI: 10.1007/978-3-030-72610-2_29
- Аннотация: Data distillation is the problem of reducing the volume of training data while keeping only the necessary information. With this paper, we deeper explore the new data distillation algorithm, previously designed for image data. Our experiments with tabular data show that the model trained on distilled samples can outperform the model trained on the original dataset. One of the problems of the considered algorithm is that produced data has poor generalization on models with different hyperparameters. We show that using multiple architectures during distillation can help overcome this problem.
- Добавил в систему: Дьяконов Александр Геннадьевич