AI Paper Atlas

Research map / Optimization / Optimizers

EvoGrad: Evolutionary-Weighted Gradient and Hessian Learning for Black-Box Optimization

Yedidya Kfir, Elad Sarafian, Yoram Louzoun, Sarit Kraus

AAAI 2026 · Optimization › Optimizers · approach: Differentiating through embedded optimization · 0 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 evolutionary (CMA covariance) weighting to gradient learningEGL · COCO suitesolved functions: 0.821 -> 0.917Improved
add Hessian (Jacobian) correction to gradient learningEGL · COCO suitesolved functions: 0.821 -> 0.858Improved
combine evolutionary weighting and Hessian correctionEvoGrad · COCO suitesolved functions: 0.917 -> 0.949Improved
replace CMA covariance weights with softmax weights (ablation)EvoGrad · COCO suitesolved functions: 0.917 -> 0.71Hurt
attach Jacobian to computation graph (full backprop) (ablation)HGrad · COCO suitesolved functions: 0.858 vs 0.858No clear effect
add trust region to CMACMA · COCO suitesolved functions: 0.612 -> 0.740Improved
vary network structure (ablation)EvoGrad2 · hyperparameter sweepcoefficient of variation: CV 0.0429No clear effect
seed EvoGrad with CMA solutionEvoGrad · Adversarial attackconvergence time: 6H -> 3HImproved
use EvoGrad instead of CMACMA · Adversarial attackMSE: 0.05 -> 0.001Improved

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