# Truly Sparse Neural Networks at Scale

arXiv: https://arxiv.org/abs/2102.01732 · HTML: https://mlsanity.com/paper/2102.01732/

## Key facts

- Cited by 16 papers in the map (data through 14 Oct 2026).
- Builds on 63 of its 106 references that are in the map.
- First posted on arXiv 2 Feb 2021.

## Builds on (most cited first, 30 of 63)

- [Communication-Efficient Learning of Deep Networks from Decentralized Data](https://mlsanity.com/paper/1602.05629/) (2016)
- [Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms](https://mlsanity.com/paper/1708.07747/) (2017)
- [Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification](https://mlsanity.com/paper/1502.01852/) (2015)
- [Deep Compression: Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffman Coding](https://mlsanity.com/paper/1510.00149/) (2015)
- [The Lottery Ticket Hypothesis: Finding Sparse, Trainable Neural Networks](https://mlsanity.com/paper/1803.03635/) (2018)
- [Learning both Weights and Connections for Efficient Neural Networks](https://mlsanity.com/paper/1506.02626/) (2015)
- [Understanding deep learning requires rethinking generalization](https://mlsanity.com/paper/1611.03530/) (2016)
- [Accurate, Large Minibatch SGD: Training ImageNet in 1 Hour](https://mlsanity.com/paper/1706.02677/) (2017)
- [Averaging Weights Leads to Wider Optima and Better Generalization](https://mlsanity.com/paper/1803.05407/) (2018)
- [Paying More Attention to Attention: Improving the Performance of Convolutional Neural Networks via Attention Transfer](https://mlsanity.com/paper/1612.03928/) (2016)
- [SNIP: Single-shot Network Pruning based on Connection Sensitivity](https://mlsanity.com/paper/1810.02340/) (2018)
- [In-Datacenter Performance Analysis of a Tensor Processing Unit](https://mlsanity.com/paper/1704.04760/) (2017)
- [Local SGD Converges Fast and Communicates Little](https://mlsanity.com/paper/1805.09767/) (2018)
- [To prune, or not to prune: exploring the efficacy of pruning for model compression](https://mlsanity.com/paper/1710.01878/) (2017)
- [Learning Sparse Neural Networks through L0 Regularization](https://mlsanity.com/paper/1712.01312/) (2017)
- [Rigging the Lottery: Making All Tickets Winners](https://mlsanity.com/paper/1911.11134/) (2019)
- [Picking Winning Tickets Before Training by Preserving Gradient Flow](https://mlsanity.com/paper/2002.07376/) (2020)
- [Sparsified SGD with Memory](https://mlsanity.com/paper/1809.07599/) (2018)
- [Scalable Training of Artificial Neural Networks with Adaptive Sparse Connectivity inspired by Network Science](https://mlsanity.com/paper/1707.04780/) (2017)
- [Parallel Restarted SGD with Faster Convergence and Less Communication: Demystifying Why Model Averaging Works for Deep Learning](https://mlsanity.com/paper/1807.06629/) (2018)
- [Predicting Parameters in Deep Learning](https://mlsanity.com/paper/1306.0543/) (2013)
- [Sparse Networks from Scratch: Faster Training without Losing Performance](https://mlsanity.com/paper/1907.04840/) (2019)
- [Deconstructing Lottery Tickets: Zeros, Signs, and the Supermask](https://mlsanity.com/paper/1905.01067/) (2019)
- [Don't Use Large Mini-Batches, Use Local SGD](https://mlsanity.com/paper/1808.07217/) (2018)
- [Gradient Sparsification for Communication-Efficient Distributed Optimization](https://mlsanity.com/paper/1710.09854/) (2017)
- [Parameter Efficient Training of Deep Convolutional Neural Networks by Dynamic Sparse Reparameterization](https://mlsanity.com/paper/1902.05967/) (2019)
- [Demystifying Parallel and Distributed Deep Learning: An In-Depth Concurrency Analysis](https://mlsanity.com/paper/1802.09941/) (2018)
- [Soft Threshold Weight Reparameterization for Learnable Sparsity](https://mlsanity.com/paper/2002.03231/) (2020)
- [QSGD: Communication-Efficient SGD via Gradient Quantization and Encoding](https://mlsanity.com/paper/1610.02132/) (2016)
- [Sparse GPU Kernels for Deep Learning](https://mlsanity.com/paper/2006.10901/) (2020)

## Built on by (most cited first, 16 of 16)

- [Ten Lessons We Have Learned in the New "Sparseland": A Short Handbook for Sparse Neural Network Researchers](https://mlsanity.com/paper/2302.02596/) (2023)
- [Fantastic Weights and How to Find Them: Where to Prune in Dynamic Sparse Training](https://mlsanity.com/paper/2306.12230/) (2023)
- [Where to Pay Attention in Sparse Training for Feature Selection?](https://mlsanity.com/paper/2211.14627/) (2022)
- [Automatic Noise Filtering with Dynamic Sparse Training in Deep Reinforcement Learning](https://mlsanity.com/paper/2302.06548/) (2023)
- [Don't Be So Dense: Sparse-to-Sparse GAN Training Without Sacrificing Performance](https://mlsanity.com/paper/2203.02770/) (2022)
- [Navigating Extremes: Dynamic Sparsity in Large Output Spaces](https://mlsanity.com/paper/2411.03171/) (2024)
- [A phase transition for finding needles in nonlinear haystacks with LASSO artificial neural networks](https://mlsanity.com/paper/2201.08652/) (2022)
- [Nerva: a Truly Sparse Implementation of Neural Networks](https://mlsanity.com/paper/2407.17437/) (2024)
- [Learning Activation Functions for Sparse Neural Networks](https://mlsanity.com/paper/2305.10964/) (2023)
- [Adaptive Sparsity Level during Training for Efficient Time Series Forecasting with Transformers](https://mlsanity.com/paper/2305.18382/) (2023)
- [Exploring the Relationship: Transformative Adaptive Activation Functions in Comparison to Other Activation Functions](https://mlsanity.com/paper/2402.09249/) (2024)
- [Are Sparse Neural Networks Better Hard Sample Learners?](https://mlsanity.com/paper/2409.09196/) (2024)
- [Sparse-to-Sparse Training of Diffusion Models](https://mlsanity.com/paper/2504.21380/) (2025)
- [NeuroTrails: Training with Dynamic Sparse Heads as the Key to Effective Ensembling](https://mlsanity.com/paper/2505.17909/) (2025)
- [Motivation is Something You Need](https://mlsanity.com/paper/2602.21064/) (2026)
- [Beyond Masked Sparsity: SNACK Enables Truly Sparse Neural Networks on GPU](https://mlsanity.com/paper/2610.04093/) (2026)

## Abstract (arXiv)

Recently, sparse training methods have started to be established as a de facto approach for training and inference efficiency in artificial neural networks. Yet, this efficiency is just in theory. In practice, everyone uses a binary mask to simulate sparsity since the typical deep learning software and hardware are optimized for dense matrix operations. In this paper, we take an orthogonal approach, and we show that we can train truly sparse neural networks to harvest their full potential. To achieve this goal, we introduce three novel contributions, specially designed for sparse neural networks: (1) a parallel training algorithm and its corresponding sparse implementation from scratch, (2) an activation function with non-trainable parameters to favour the gradient flow, and (3) a hidden neurons importance metric to eliminate redundancies. All in one, we are able to break the record and to train the largest neural network ever trained in terms of representational power -- reaching the bat brain size. The results show that our approach has state-of-the-art performance while opening the path for an environmentally friendly artificial intelligence era.