AI Paper Atlas

Research map / Optimization / Optimizers

Binomial Gradient-Based Meta-Learning for Enhanced Meta-Gradient Estimation

Yilang Zhang, Abraham Jaeger Mountain, Bingcong Li, Georgios B. Giannakis

ICLR 2026 · Optimization › Optimizers · approach: Gradient-based meta-learning theory · 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
use binomial expansion instead of truncated backprop (L=1)TruncMAML · miniImageNet 1-shotaccuracy (%): 44.53 -> 45.50Improved
use binomial expansion instead of truncated backprop (L=2)TruncMAML · miniImageNet 1-shotaccuracy (%): 44.93 -> 46.23Improved
use binomial expansion instead of truncated backprop (L=3)TruncMAML · miniImageNet 1-shotaccuracy (%): 44.33 -> 46.00Improved
use binomial expansion instead of truncated backprop (L=1)TruncMAML · tieredImageNet 1-shotaccuracy (%): 43.67 -> 46.23Improved
use binomial expansion instead of truncated backprop (L=1 and 2)TruncMAML · miniImageNet 5-shotaccuracy (%): 62.43 -> 62.36 (L=1), 63.61 -> 63.49 (L=2)Mixed
use binomial expansion instead of truncated backprop, averaged over LTruncMAML · 1-shot vs 5-shotaverage accuracy gap: +1.33 average (1-shot), +0.27 average (5-shot)Mixed
use BinomMAML instead of iMAML (L=4)iMAML · miniImageNet 1-shotaccuracy (%): 45.43 -> 46.24Improved
use BinomMAML instead of iMAML (L=4)iMAML · tieredImageNet 5-shotaccuracy (%): 61.80 -> 64.63Improved
increase truncation L from 1 to 4 for BinomMAML (ablation)BinomMAML · miniImageNet 1-shotaccuracy (%): 45.50 -> 46.24Improved
increase truncation L from 1 to 4 for TruncMAML (ablation)TruncMAML · miniImageNet 1-shotaccuracy (%): 44.53 -> 44.83No clear effect
increase truncation L from 1 to 4 for BinomMAML (ablation)BinomMAML · tieredImageNet 5-shotaccuracy (%): 64.49 -> 64.63No clear effect
BinomMAML versus truncated backprop in meta-gradient error (L=4)TruncMAML · synthetic sinusoid regressionmeta-gradient errorImproved

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