AI Paper Atlas

Research map / Architectures and efficiency / Neural architecture search

Resource Constrained Neural Network Architecture Search: Will a Submodularity Assumption Help?

Yunyang Xiong, Ronak Mehta, Vikas Singh

ICCV 2019 · Architectures and efficiency › Neural architecture search · approach: Bi-level differentiable NAS · 36 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add block diversity (search over all six block types instead of only type 5) (ablation)RCNet · CIFAR-100top-1 accuracy: 75.8 -> 76.1Improved
use the greedy search procedure instead of random blocks for the last several blocks (ablation)RCNet · CIFAR-100top-1 accuracy: 74.9 -> 76.1Improved
randomly add one more block to the searched architecture (ablation)RCNet · CIFAR-100top-1 accuracy: 76.1 -> 75.1Hurt
add accumulated random blocks (diminishing returns of accuracy with early stopping)RCAS set function · CIFAR-100top-1 accuracy: 56.96% -> 60.72% over eight additions, gains shrinkingImproved
RCNet instead of MobileNetV2 on CIFAR-100RCNet · CIFAR-100top-1 accuracy: 74.2 -> 76.1Improved
RCNet instead of DARTS on CIFAR-100RCNet · CIFAR-100top-1 accuracy: 75.9 -> 76.1Improved
RCNet instead of MobileNetV2 on ImageNetRCNet · ImageNet valtop-1 accuracy: 72.0 -> 72.2Improved
RCNet-B instead of DARTS on ImageNetRCNet-B · ImageNet valtop-1 accuracy: 73.1 -> 74.7Improved
RCNet-B versus MNasNet-92 on ImageNetRCNet-B · ImageNet valtop-1 accuracy: 74.8 -> 74.7No clear effect
transfer the CIFAR-100 searched architecture to ImageNetRCNet · ImageNet valtop-1 accuracyNo clear effect
RCNet instead of MobileNetV2 on iPhone 5sRCNet · iPhone 5s CoreMLinference time (ms): 197.2 -> 183.5No clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy