Open Peer Review. Open Publishing. Open Access. Open Discussion. Open Directory. Open Recommendations. Open API. Open Source.
Bayesian Incremental Learning for Deep Neural Networks
Max Kochurov, Timur Garipov, Dmitry Podoprikhin, Dmitry Molchanov, Arsenii Ashukha, Dmitry Vetrov
Feb 12, 2018 (modified: Mar 27, 2018)ICLR 2018 Workshop Submissionreaders: everyone
Abstract: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.
TL;DR:We propose a Bayesian incremental learning algorithm with a way to use pre-trained DNNs