Structured Ensembles: an Approach to Reduce the Memory Footprint of Ensemble Methods
This addresses the memory footprint problem for practitioners using ensemble methods in deep learning, offering an incremental improvement over existing approaches.
The paper tackles the high memory cost of ensemble methods in deep neural networks by proposing Structured Ensembles, which extracts multiple sub-networks from a single untrained network through end-to-end optimization, achieving higher or comparable accuracy with significantly less storage and favorable predictive calibration and uncertainty compared to state-of-the-art.
In this paper, we propose a novel ensembling technique for deep neural networks, which is able to drastically reduce the required memory compared to alternative approaches. In particular, we propose to extract multiple sub-networks from a single, untrained neural network by solving an end-to-end optimization task combining differentiable scaling over the original architecture, with multiple regularization terms favouring the diversity of the ensemble. Since our proposal aims to detect and extract sub-structures, we call it Structured Ensemble. On a large experimental evaluation, we show that our method can achieve higher or comparable accuracy to competing methods while requiring significantly less storage. In addition, we evaluate our ensembles in terms of predictive calibration and uncertainty, showing they compare favourably with the state-of-the-art. Finally, we draw a link with the continual learning literature, and we propose a modification of our framework to handle continuous streams of tasks with a sub-linear memory cost. We compare with a number of alternative strategies to mitigate catastrophic forgetting, highlighting advantages in terms of average accuracy and memory.