Anytime Inference with Distilled Hierarchical Neural Ensembles

TL;DR

This paper introduces Hierarchical Neural Ensembles (HNE), a framework enabling flexible anytime inference by sharing layers in a tree structure, and a hierarchical distillation method to improve small ensemble accuracy, achieving state-of-the-art results.

Contribution

The paper presents a novel hierarchical ensemble framework and a distillation technique that enhance anytime inference efficiency and accuracy in deep neural networks.

Findings

HNE achieves superior accuracy-computation trade-offs on CIFAR-10/100 and ImageNet.

Hierarchical distillation improves small ensemble performance.

HNE allows dynamic control of inference complexity.

Abstract

Inference in deep neural networks can be computationally expensive, and networks capable of anytime inference are important in mscenarios where the amount of compute or quantity of input data varies over time. In such networks the inference process can interrupted to provide a result faster, or continued to obtain a more accurate result. We propose Hierarchical Neural Ensembles (HNE), a novel framework to embed an ensemble of multiple networks in a hierarchical tree structure, sharing intermediate layers. In HNE we control the complexity of inference on-the-fly by evaluating more or less models in the ensemble. Our second contribution is a novel hierarchical distillation method to boost the prediction accuracy of small ensembles. This approach leverages the nested structure of our ensembles, to optimally allocate accuracy and diversity across the individual models. Our experiments show that, compared to previous anytime inference models, HNE provides state-of-the-art accuracy-computate trade-offs on the CIFAR-10/100 and ImageNet datasets.