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

Download:

Current browse context:

cs.CV

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

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

Computer Science > Computer Vision and Pattern Recognition

Title: Estimating and Maximizing Mutual Information for Knowledge Distillation

Abstract: In this work, we propose Mutual Information Maximization Knowledge Distillation (MIMKD). Our method uses a contrastive objective to simultaneously estimate and maximize a lower bound on the mutual information of local and global feature representations between a teacher and a student network. We demonstrate through extensive experiments that this can be used to improve the performance of low capacity models by transferring knowledge from more performant but computationally expensive models. This can be used to produce better models that can be run on devices with low computational resources. Our method is flexible, we can distill knowledge from teachers with arbitrary network architectures to arbitrary student networks. Our empirical results show that MIMKD outperforms competing approaches across a wide range of student-teacher pairs with different capacities, with different architectures, and when student networks are with extremely low capacity. We are able to obtain 74.55% accuracy on CIFAR100 with a ShufflenetV2 from a baseline accuracy of 69.8% by distilling knowledge from ResNet-50. On Imagenet we improve a ResNet-18 network from 68.88% to 70.32% accuracy (1.44%+) using a ResNet-34 teacher network.
Subjects: Computer Vision and Pattern Recognition (cs.CV); Information Theory (cs.IT)
Cite as: arXiv:2110.15946 [cs.CV]
  (or arXiv:2110.15946v2 [cs.CV] for this version)

Submission history

From: Aman Shrivastava [view email]
[v1] Fri, 29 Oct 2021 17:49:56 GMT (1183kb,D)
[v2] Mon, 29 Nov 2021 18:24:36 GMT (1704kb,D)

Link back to: arXiv, form interface, contact.