AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

A fast heuristic to optimize time-space tradeoff for large models

Akifumi Imanishi, Zijian Xu, Masayuki Takagi, Sixue Wang, Emilio Castillo

NeurIPS 2023 · Architectures and efficiency › RNNs and normalization · approach: Efficient network compression and training · 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 FastSA instead of Moccasinrecomputation planning · RL 2 graph, 80% budgetcost increase: 4.9 -> 0.0Improved
use FastSA instead of Moccasinrecomputation planning · RL 4 graph, 80% budgetcost increase: 3.4 -> 2.5Improved
use FastSA instead of Checkmate MILPrecomputation planning · RL 1 graph, 80% budgetcost increase: 2.3 -> 0.3Improved
use FastSA instead of Checkmate MILPrecomputation planning · CM 2 graph, 80% budgetcost increase: 0.3 -> 0.4No clear effect
use FastSA instead of Checkmate MILPrecomputation planning · ResNet18, 80% budgetcost increase: 1.8 -> 2.9Hurt
use FastSA instead of Checkmate MILPrecomputation planning · ResNet18, 50% budgetsolver time: up to 1000x fasterImproved
use FastSA instead of Checkmate MILPrecomputation planning · VGG11, 80% budgetcost increase: 2.9 -> 2.9No clear effect
use FastSA instead of Checkmate LPrecomputation planning · 23 vision/text modelsmemory reduction: 73%Improved
tighten memory budget from 50% to 25%FastSA · 23 vision/text modelsaverage execution time increase: 7% -> 18%Hurt
add node grouping before simulated annealing (ablation)FastSA · LLaMA, 10% memory budgetfinal objective value: 4.4 times larger without groupingImproved
use node grouping alone without SA (ablation)FastSA · LLaMA, 10% memory budgetpeak memory: 3.5e+11 -> 7.8e+10Improved
add 70% budget setting for FastSAFastSA · RL 4 graphcost increase: 2.5 -> 7.4Hurt

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