Papers › Do We Actually Need Dense Over-Parameterization? In-Time Over-Parameterization in...

Do We Actually Need Dense Over-Parameterization? In-Time Over-Parameterization in Sparse Training

4 Feb 2021arXiv:2102.02887archive 2025-07-28

Shiwei Liu, Lu Yin, Decebal Constantin Mocanu, Mykola Pechenizkiy

In this paper, we introduce a new perspective on training deep neural networks capable of state-of-the-art performance without the need for the expensive over-parameterization by proposing the concept of In-Time Over-Parameterization (ITOP) in sparse training. By starting from a random sparse network and continuously exploring sparse connectivities during training, we can perform an Over-Parameterization in the space-time manifold, closing the gap in the expressibility between sparse training and dense training. We further use ITOP to understand the underlying mechanism of Dynamic Sparse Training (DST) and indicate that the benefits of DST come from its ability to consider across time all possible parameters when searching for the optimal sparse connectivity. As long as there are sufficient parameters that have been reliably explored during training, DST can outperform the dense neural network by a large margin. We present a series of experiments to support our conjecture and achieve the state-of-the-art sparse training performance with ResNet-50 on ImageNet. More impressively, our method achieves dominant performance over the overparameterization-based sparse methods at extreme sparsity levels. When trained on CIFAR-100, our method can match the performance of the dense model even at an extreme sparsity (98%). Code can be found https://github.com/Shiweiliuiiiiiii/In-Time-Over-Parameterization.

PaperPDFCode

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

Shiweiliuiiiiiii/GraNet mentioned on GitHubpytorch report
stevenboys/agent mentioned on GitHubpytorch report
vita-group/granet mentioned on GitHubpytorch report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Image ClassificationSparse Learning

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Sparse Learning ImageNet Resnet-50: 80% Sparse 100 epochs Top-1 Accuracy 75.84 #4 of 9 Archive leaderboard report
Sparse Learning ImageNet Resnet-50: 90% Sparse 100 epochs Top-1 Accuracy 73.82 #6 of 9 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

DST

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections