AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

AutoDV: An End-to-End Deep Learning Model for High-Dimensional Data Visualization

Wei Dai, Jicong Fan

ICLR 2026 · Architectures and efficiency › RNNs and normalization · approach: Differentiable optimization layers

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 5 multi-scale graphs instead of 1 (ablation)AutoDV-tSNE · Image graph-count sensitivity (5 train datasets, 1 test)test NMI (%): 59.83 -> 82.48Improved
increase graphs from 5 to 9 (ablation)AutoDV-tSNE · Image graph-count sensitivity (5 train datasets, 1 test)test NMI (%): 82.48 -> 53.27Hurt
increase graphs from 5 to 7 (ablation)AutoDV-tSNE · Image graph-count sensitivity (5 train datasets, 1 test)test NMI (%): 82.48 -> 74.08Hurt
increase PE dimension from 8 to 64 (ablation)AutoDV-tSNE · Image PE-dim sensitivity (5 train datasets, 1 test)test NMI (%): 78.41 -> 82.48Improved
increase PE dimension from 64 to 128 (ablation)AutoDV-tSNE · Image PE-dim sensitivity (5 train datasets, 1 test)test NMI (%): 82.48 -> 73.24Hurt
use full UMAP-style loss instead of local-only loss (ablation)AutoDV-UMAP · CIFAR10 test subsets (trained on MNIST/FMNIST)NMI (%)Improved
use sign-count (l0) PE sign flipping instead of random or default l-infinity (ablation)AutoDV-tSNE · CIFAR10 test subsets (trained on MNIST/FMNIST)NMI (%)Improved
train AutoDV on historical t-SNE embeddings and test on unseen tabular dataAutoDV-tSNE · UCI tabular test subsets (156 subsets)NMI (%): 30.92 -> 33.45Improved
train AutoDV on historical t-SNE embeddings and test on unseen gene dataAutoDV-tSNE · Mouse Retina gene test subsetsNMI (%): 32.73 -> 33.22Improved
apply AutoDV to unseen CIFAR10 instead of tuned t-SNEAutoDV-tSNE · CIFAR10 test subsets (trained on MNIST/FMNIST)NMI (%): 77.04 -> 68.70Hurt

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