AI Paper Atlas

Research map / Architectures and efficiency / Neural architecture search

Multinomial Distribution Learning for Effective Neural Architecture Search

Xiawu Zheng, Rongrong Ji, Lang Tang, Baochang Zhang, Jianzhuang Liu, Qi Tian

ICCV 2019 · Architectures and efficiency › Neural architecture search · approach: Efficient NAS search strategies · 123 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
use MdeNAS multinomial distribution search instead of random sampling in the same spacerandom sampling · CIFAR-10test error (%): 3.49 -> 2.55Improved
use MdeNAS instead of NASNet-A searchNASNet-A · CIFAR-10search cost (GPU days): 1800 -> 0.16Improved
use MdeNAS instead of DARTS second orderDARTS · CIFAR-10test error (%): 2.83 -> 2.55Improved
use MdeNAS instead of Path-level NASPath-level NAS · CIFAR-10test error (%): 2.49 -> 2.55Hurt
search directly on ImageNet with MdeNAS instead of Proxyless (GPU)Proxyless (GPU) · ImageNet mobile (searched directly on ImageNet)top-1 accuracy (%): 74.8 -> 75.2Improved
search directly on ImageNet with MdeNAS instead of Proxyless (CPU)Proxyless (CPU) · ImageNet mobile (searched directly on ImageNet)top-1 accuracy (%): 74.1 -> 74.1No clear effect
search directly on ImageNet with MdeNAS (GPU) instead of Proxyless (GPU) latencyProxyless (GPU) · ImageNet mobile (searched directly on ImageNet)GPU latency (ms): 5.1ms -> 4.9msImproved
transfer CIFAR-10 cell instead of AmoebaNet-CAmoebaNet-C · ImageNet mobile (searched on CIFAR-10, transferred)top-1 accuracy (%): 75.7 -> 74.5Hurt
evaluate architectures early in training (ranking hypothesis)architecture ranking · CIFAR-10Kendall tau: 0.474Improved

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