AI Paper Atlas

Research map / Architectures and efficiency / Neural architecture search

HyTAS: A Hyperspectral Image Transformer Architecture Search Benchmark and Analysis

Fangqin Zhou, Mert Kilickaya, Joaquin Vanschoren, Ran Piao

ECCV 2024 · Architectures and efficiency › Neural architecture search · approach: Bi-level differentiable NAS · 2 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 layer decay to ZiCo's score aggregation and use Fisher-style activation gradients (ZiCo++)ZiCo · Indian PinesSpearman correlation: 0.52 -> 0.73Improved
ZiCo++ instead of ZiCoproxy search · PaviaUtest OA: 0.84 -> 0.84No clear effect
ZiCo++ instead of ZiCoproxy search · Salinastest OA: 0.91 -> 0.94Improved
use log(msa x mlp) instead of msa + mlp score (ablation)Synflow · Indian PinesSpearman correlation: 0.69 -> 0.72Improved
use log(msa x mlp) instead of msa + mlp score (ablation)Synflow · PaviaUSpearman correlation: 0.69 -> 0.63Hurt
use log(msa x mlp) instead of msa + mlp score (ablation)SNIP · five HSI datasetsSpearman correlation: 0.70 -> 0.70 (Indian Pines)No clear effect
use random input batch instead of real HSI data for proxies (ablation)data-dependent proxies · Indian Pinestest OA / Spearman correlation: GradNorm 0.81 -> 0.81No clear effect
random input for JacobCov (ablation)JacobCov · Indian PinesSpearman correlation: 0.58 -> 0.68 (+0.1)Improved
remove sign operations from Synflow, LogSynflow and DSS (ablation)Synflow/LogSynflow/DSS · Indian Pinessearch performanceImproved
train Random Forest on architecture features plus proxy scores to predict OAproxy-based ranking · Indian PinesSpearman correlation: 0.80 (9% above best proxy)Improved
constrain search to larger model sizes (ablation)proxies · Indian PinesSpearman correlationHurt
choose proxies' top architecture instead of human-designed SpectralFormerHSI transformer design · Indian Pinestest OA: 0.79 -> 0.81Improved

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