# Automated Pruning for Deep Neural Network Compression

arXiv: https://arxiv.org/abs/1712.01721 · HTML: https://mlsanity.com/paper/1712.01721/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 5 Dec 2017.

## Built on by (most cited first, 6 of 6)

- [Compacting Deep Neural Networks for Internet of Things: Methods and Applications](https://mlsanity.com/paper/2103.11083/) (2021)
- [SD-Conv: Towards the Parameter-Efficiency of Dynamic Convolution](https://mlsanity.com/paper/2204.02227/) (2022)
- [Optimizing Dense Feed-Forward Neural Networks](https://mlsanity.com/paper/2312.10560/) (2023)
- [Cascade Weight Shedding in Deep Neural Networks: Benefits and Pitfalls for Network Pruning](https://mlsanity.com/paper/2103.10629/) (2021)
- [Knowledge Distillation By Sparse Representation Matching](https://mlsanity.com/paper/2103.17012/) (2021)
- [Solving hybrid machine learning tasks by traversing weight space geodesics](https://mlsanity.com/paper/2106.02793/) (2021)

## Abstract (arXiv)

In this work we present a method to improve the pruning step of the current state-of-the-art methodology to compress neural networks. The novelty of the proposed pruning technique is in its differentiability, which allows pruning to be performed during the backpropagation phase of the network training. This enables an end-to-end learning and strongly reduces the training time. The technique is based on a family of differentiable pruning functions and a new regularizer specifically designed to enforce pruning. The experimental results show that the joint optimization of both the thresholds and the network weights permits to reach a higher compression rate, reducing the number of weights of the pruned network by a further 14% to 33% compared to the current state-of-the-art. Furthermore, we believe that this is the first study where the generalization capabilities in transfer learning tasks of the features extracted by a pruned network are analyzed. To achieve this goal, we show that the representations learned using the proposed pruning methodology maintain the same effectiveness and generality of those learned by the corresponding non-compressed network on a set of different recognition tasks.