# ParaDiS: Parallelly Distributable Slimmable Neural Networks

arXiv: https://arxiv.org/abs/2110.02724 · HTML: https://mlsanity.com/paper/2110.02724/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 21 of its 37 references that are in the map.
- First posted on arXiv 6 Oct 2021.

## Builds on (most cited first, 21 of 21)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [Distilling the Knowledge in a Neural Network](https://mlsanity.com/paper/1503.02531/) (2015)
- [Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift](https://mlsanity.com/paper/1502.03167/) (2015)
- [MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications](https://mlsanity.com/paper/1704.04861/) (2017)
- [Self-training with Noisy Student improves ImageNet classification](https://mlsanity.com/paper/1911.04252/) (2019)
- [ShuffleNet: An Extremely Efficient Convolutional Neural Network for Mobile Devices](https://mlsanity.com/paper/1707.01083/) (2017)
- [Once-for-All: Train One Network and Specialize it for Efficient Deployment](https://mlsanity.com/paper/1908.09791/) (2019)
- [Weight Normalization: A Simple Reparameterization to Accelerate Training of Deep Neural Networks](https://mlsanity.com/paper/1602.07868/) (2016)
- [Slimmable Neural Networks](https://mlsanity.com/paper/1812.08928/) (2018)
- [Universally Slimmable Networks and Improved Training Techniques](https://mlsanity.com/paper/1903.05134/) (2019)
- [BigNAS: Scaling Up Neural Architecture Search with Big Single-Stage Models](https://mlsanity.com/paper/2003.11142/) (2020)
- [Distributed Deep Neural Networks over the Cloud, the Edge and End Devices](https://mlsanity.com/paper/1709.01921/) (2017)
- [Split Computing and Early Exiting for Deep Learning Applications: Survey and Research Challenges](https://mlsanity.com/paper/2103.04505/) (2021)
- [Dynamic Slimmable Network](https://mlsanity.com/paper/2103.13258/) (2021)
- [Ensemble Knowledge Distillation for Learning Improved and Efficient Networks](https://mlsanity.com/paper/1909.08097/) (2019)
- [CompOFA: Compound Once-For-All Networks for Faster Multi-Platform Deployment](https://mlsanity.com/paper/2104.12642/) (2021)
- [Memory- and Communication-Aware Model Compression for Distributed Deep Learning Inference on IoT](https://mlsanity.com/paper/1907.11804/) (2019)
- [Context-aware Deep Feature Compression for High-speed Visual Tracking](https://mlsanity.com/paper/1803.10537/) (2018)
- [Anytime Inference with Distilled Hierarchical Neural Ensembles](https://mlsanity.com/paper/2003.01474/) (2020)
- [Collaborative Execution of Deep Neural Networks on Internet of Things Devices](https://mlsanity.com/paper/1901.02537/) (2019)
- [Inplace knowledge distillation with teacher assistant for improved training of flexible deep neural networks](https://mlsanity.com/paper/2105.08369/) (2021)

## Abstract (arXiv)

When several limited power devices are available, one of the most efficient ways to make profit of these resources, while reducing the processing latency and communication load, is to run in parallel several neural sub-networks and to fuse the result at the end of processing. However, such a combination of sub-networks must be trained specifically for each particular configuration of devices (characterized by number of devices and their capacities) which may vary over different model deployments and even within the same deployment. In this work we introduce parallelly distributable slimmable (ParaDiS) neural networks that are splittable in parallel among various device configurations without retraining. While inspired by slimmable networks allowing instant adaptation to resources on just one device, ParaDiS networks consist of several multi-device distributable configurations or switches that strongly share the parameters between them. We evaluate ParaDiS framework on MobileNet v1 and ResNet-50 architectures on ImageNet classification task and WDSR architecture for image super-resolution task. We show that ParaDiS switches achieve similar or better accuracy than the individual models, i.e., distributed models of the same structure trained individually. Moreover, we show that, as compared to universally slimmable networks that are not distributable, the accuracy of distributable ParaDiS switches either does not drop at all or drops by a maximum of 1 % only in the worst cases. Finally, once distributed over several devices, ParaDiS outperforms greatly slimmable models.