Bayesian Incremental Learning for Deep Neural Networks

In industrial machine learning pipelines, data often arrive in parts. Particularly in the case of deep neural networks, it may be too expensive to train the model from scratch each time, so one would rather use a previously learned model and the new data to improve performance. However, deep neural...

Full description

Saved in:
Bibliographic Details
Published in:arXiv.org 2018-03
Main Authors: Kochurov, Max, Garipov, Timur, Podoprikhin, Dmitry, Molchanov, Dmitry, Ashukha, Arsenii, Vetrov, Dmitry
Format: Article
Language:English
Subjects:
Bayesian analysis
Case depth
Machine learning
Neural networks
Performance enhancement
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In industrial machine learning pipelines, data often arrive in parts. Particularly in the case of deep neural networks, it may be too expensive to train the model from scratch each time, so one would rather use a previously learned model and the new data to improve performance. However, deep neural networks are prone to getting stuck in a suboptimal solution when trained on only new data as compared to the full dataset. Our work focuses on a continuous learning setup where the task is always the same and new parts of data arrive sequentially. We apply a Bayesian approach to update the posterior approximation with each new piece of data and find this method to outperform the traditional approach in our experiments.
ISSN:2331-8422