Summary of Sae: Single Architecture Ensemble Neural Networks, by Martin Ferianc et al.
SAE: Single Architecture Ensemble Neural Networks
by Martin Ferianc, Hongxiang Fan, Miguel Rodrigues
First submitted to arxiv on: 9 Feb 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The Single Architecture Ensemble (SAE) framework is a novel approach that enables automatic and joint search through early exit and multi input multi output configurations, along with their previously unobserved in-between combinations. This framework consists of two parts: a scalable search space that generalises previous methods, and an optimisation objective that learns the optimal configuration for a given task. SAE is shown to be effective in image classification and regression experiments, achieving competitive accuracy or confidence calibration to baselines while reducing compute operations or parameter count by up to 1.5-3.7 times. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Ensembles of separate neural networks (NNs) have been found to outperform single NNs in various tasks. To make these ensembles more efficient on hardware, researchers have developed methods like adding early exits or considering multi input multi output approaches. However, it is unclear which method works best for a given task, requiring a manual search through each method. The SAE framework solves this problem by automatically searching through different configurations and finding the optimal one for a specific task. |
Keywords
* Artificial intelligence * Image classification * Regression
