AI Paper Atlas

Research map / Federated learning and privacy / Communication-efficient FL

Interlocking Backpropagation: Improving depthwise model-parallelism

Aidan N. Gomez, Oscar Key, Kuba Perlin, Stephen Gou, Nick Frosst, Jeff Dean, Yarin Gal

JMLR 2022 · Federated learning and privacy › Communication-efficient FL · approach: Decentralized peer-to-peer training · 9 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 2-wise interlocking to local (1-wise) training (ablation)ResNet-32 · CIFAR-100 testaccuracy (%): 75.02 -> 78.09Improved
add 2-wise interlocking to local (1-wise) training (ablation)ResNet-50 · ImageNet valaccuracy (%): 72.05 -> 74.45Improved
increase interlocking from 2-wise to 3-wise (ablation)ResNet-50 · ImageNet valaccuracy (%): 74.45 -> 76.27Improved
increase interlocking from 2-wise to 3-wise (ablation)ResNet-32 · CIFAR-100 testaccuracy (%): 78.09 -> 77.84No clear effect
replace end-to-end training with 2-wise interlockingResNet-32 · CIFAR-100 testaccuracy (%): 76.71 -> 78.09Improved
replace end-to-end training with 3-wise interlockingResNet-32 · CIFAR-10 testaccuracy (%): 95.20 -> 95.42Improved
use 2-wise interlocking instead of end-to-end under fixed time budgetResNet-32 · CIFAR-100 test (fixed time budget)accuracy (%): 73.98 -> 76.85Improved
use 2-wise interlocking instead of local 1-wise under fixed time budgetResNet-32 · CIFAR-100 test (fixed time budget)accuracy (%): 74.33 -> 76.85Improved
use fine-tuning instead of 2-wise interlocking under fixed time budgetResNet-32 · CIFAR-10 test (fixed time budget)accuracy (%): 94.06 -> 94.21No clear effect
ensemble predictions of auxiliary heads across modules (ablation)interlocking backprop · CIFAR-10 / ensemblingaccuracyNo clear effect
use 2-wise interlocking instead of end-to-end, fixed number of stepsTransformer · One Billion Word (4 modules, fixed steps)test perplexityHurt
use 2-wise interlocking instead of end-to-end, per-time comparisonTransformer · One Billion Word (4 modules, 8 days)test perplexity: less than half the training step time for similar test perplexityImproved

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