We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations

DBLP - CS Bibliography


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Machine Learning

Title: High-dimensional Bayesian Optimization for CNN Auto Pruning with Clustering and Rollback

Abstract: Pruning has been widely used to slim convolutional neural network (CNN) models to achieve a good trade-off between accuracy and model size so that the pruned models become feasible for power-constrained devices such as mobile phones. This process can be automated to avoid the expensive hand-crafted efforts and to explore a large pruning space automatically so that the high-performance pruning policy can be achieved efficiently. Nowadays, reinforcement learning (RL) and Bayesian optimization (BO)-based auto pruners are widely used due to their solid theoretical foundation, universality, and high compressing quality. However, the RL agent suffers from long training times and high variance of results, while the BO agent is time-consuming for high-dimensional design spaces. In this work, we propose an enhanced BO agent to obtain significant acceleration for auto pruning in high-dimensional design spaces. To achieve this, a novel clustering algorithm is proposed to reduce the dimension of the design space to speedup the searching process. Then, a roll-back algorithm is proposed to recover the high-dimensional design space so that higher pruning accuracy can be obtained. We validate our proposed method on ResNet, MobileNet, and VGG models, and our experiments show that the proposed method significantly improves the accuracy of BO when pruning very deep CNN models. Moreover, our method achieves lower variance and shorter time than the RL-based counterpart.
Comments: 7 pages with 1 page for references
Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
Cite as: arXiv:2109.10591 [cs.LG]
  (or arXiv:2109.10591v1 [cs.LG] for this version)

Submission history

From: Jiandong Mu [view email]
[v1] Wed, 22 Sep 2021 08:39:15 GMT (4671kb,D)

Link back to: arXiv, form interface, contact.