# Dynamically Sacrificing Accuracy for Reduced Computation: Cascaded Inference Based on Softmax Confidence

arXiv: https://arxiv.org/abs/1805.10982 · HTML: https://mlsanity.com/paper/1805.10982/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 28 May 2018.

## Built on by (most cited first, 5 of 5)

- [EBJR: Energy-Based Joint Reasoning for Adaptive Inference](https://mlsanity.com/paper/2110.10343/) (2021)
- [Early-exit Convolutional Neural Networks](https://mlsanity.com/paper/2409.05336/) (2024)
- [Energy-Aware Deep Learning on Resource-Constrained Hardware](https://mlsanity.com/paper/2505.12523/) (2025)
- [PTEENet: Post-Trained Early-Exit Neural Networks Augmentation for Inference Cost Optimization](https://mlsanity.com/paper/2501.02508/) (2025)
- [Compact Multi-level Sparse Neural Networks with Input Independent Dynamic Rerouting](https://mlsanity.com/paper/2112.10930/) (2021)

## Abstract (arXiv)

We study the tradeoff between computational effort and classification accuracy in a cascade of deep neural networks. During inference, the user sets the acceptable accuracy degradation which then automatically determines confidence thresholds for the intermediate classifiers. As soon as the confidence threshold is met, inference terminates immediately without having to compute the output of the complete network. Confidence levels are derived directly from the softmax outputs of intermediate classifiers, as we do not train special decision functions. We show that using a softmax output as a confidence measure in a cascade of deep neural networks leads to a reduction of 15%-50% in the number of MAC operations while degrading the classification accuracy by roughly 1%. Our method can be easily incorporated into pre-trained non-cascaded architectures, as we exemplify on ResNet. Our main contribution is a method that dynamically adjusts the tradeoff between accuracy and computation without retraining the model.