AI Paper Atlas

Research map / Architectures and efficiency / Neural architecture search

IDARTS: Interactive Differentiable Architecture Search

Song Xue, Runqi Wang, Baochang Zhang, Tian Wang, Guodong Guo, David Doermann

ICCV 2021 · Architectures and efficiency › Neural architecture search · approach: Bi-level differentiable NAS · 14 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 backtracking of architecture parameters (IDARTS) on top of PC-DARTS-style searchPC-DARTS · CIFAR10 testaccuracy (%): 97.43 -> 97.68Improved
add backtracking of architecture parameters (IDARTS) on top of PC-DARTS-style searchPC-DARTS · ImageNet val (searched on ImageNet)top-1 accuracy (%): 75.8 -> 76.52Improved
add backtracking of architecture parameters (IDARTS)PC-DARTS · ImageNet val (searched on CIFAR10)top-1 accuracy (%): 74.9 -> 76.14Improved
IDARTS versus CDARTS (cyclic feedback search)CDARTS · CIFAR10 testaccuracy (%): 97.52 -> 97.68Improved
IDARTS versus CDARTS search costCDARTS · CIFAR10 testsearch cost (GPU days): 0.3 -> 0.1Improved
search directly on ImageNet instead of transferring a CIFAR10-searched cellIDARTS · ImageNet valtop-1 accuracy (%): 76.14 -> 76.52Improved
search directly on ImageNet instead of CIFAR10 (search cost)IDARTS · ImageNet valsearch cost (GPU days): 0.1 -> 3.8Hurt
IDARTS versus ProxylessNASProxylessNAS · CIFAR10 testaccuracy (%): 97.92 -> 97.68Hurt
IDARTS versus manual DenseNet-BCDenseNet-BC · CIFAR10 testaccuracy (%): 96.54 -> 97.68Improved
IDARTS versus manual ShuffleNet 2x (v2)ShuffleNet 2x (v2) · ImageNet valtop-1 accuracy (%): 74.9 -> 76.52Improved
add backtracking so alpha escapes local minima during search (ablation)PC-DARTS · CIFAR10 searchfinal search validation lossImproved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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