AI Paper Atlas

Research map / Optimization / Optimizers

Improving Neural Network Training in Low Dimensional Random Bases

Frithjof Gressmann, Zach Eaton-Rosen, Carlo Luschi

NeurIPS 2020 · Optimization › Optimizers · approach: Randomized gradient estimation · 8 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
Re-draw the random subspace at every step (RBD) instead of a fixed random projection (FPD) (ablation)random subspace training d=250 · CNN-CIFAR10validation accuracy (%): 37.2 -> 54.64Improved
Re-draw the random subspace at every step (RBD) instead of a fixed projection (ablation)random subspace training d=250 · FC-CIFAR10validation accuracy (%): 21.4 -> 43.77Improved
Re-draw the random subspace at every step (RBD) instead of a fixed projection (ablation)random subspace training d=250 · FC-MNISTvalidation accuracy (%): 80 -> 93.893Improved
Use RBD instead of evolution-strategy black-box optimization (NES) at same dimensionality (ablation)random subspace training d=250 · FC-MNISTvalidation accuracy (%): 22.5 -> 93.893Improved
Increase the number of random directions (reduction factor from 75x to 10x) (ablation)RBD · reduced CIFAR-10 networkvalidation accuracy (%): 62.44 -> 70.26Improved
Train longer (2000 epochs) with only d=2 directions (ablation)RBD · CNN-CIFAR10validation accuracy (%): 68.06 -> 48.20Hurt
Use more directions (10 and 50) instead of 2 at convergence (ablation)RBD · CNN-CIFAR10validation accuracy (%): 48.20 -> higher than 2-direction limitImproved
Replace Gaussian bases with Uniform bases (ablation)RBD · CNN-CIFAR10validation accuracy (%): 52.3 -> 41Hurt
Replace Gaussian bases with Bernoulli-0.5 bases (ablation)RBD · CNN-CIFAR10validation accuracy (%): 52.3 -> 28.5Hurt
Replace Gaussian bases with Bernoulli-0.5 bases (ablation)RBD · FC-MNISTvalidation accuracy (%): 94.95 -> 77.5Hurt
Switch between RBD and SGD during training without tuning learning rates (ablation)RBD/SGD hybrid · CIFAR10-CNNvalidation accuracyNo clear effect
Start with SGD then switch to RBD (ablation)RBD after SGD · CIFAR10-CNNvalidation accuracyHurt

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