AI Paper Atlas

Research map / Learning theory and algorithms / Prediction and decision losses

Machine learning detects terminal singularities

Tom Coates, Alexander M. Kasprzyk, Sara Veneziale

NeurIPS 2023 · Learning theory and algorithms › Prediction and decision losses · approach: Theory of neural network sensitivity · 2 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
test on weight matrices with entries bounded by 8 instead of 7MLP terminality classifier · Terminality classificationaccuracy: 95% -> 62%Hurt
test on weight matrices with entries bounded by 10 instead of 7MLP terminality classifier · Terminality classificationaccuracy: 95% -> 50%Hurt
test on entries bounded by 9 instead of 7MLP terminality classifier · Terminality classificationaccuracy: 95% -> 52%Hurt
retrain on larger weight bound ten with 5M samples (ablation)MLP terminality classifier · Terminality classificationaccuracy: 95% -> 90%Hurt
increase training set from 5M to 10M for bound ten (ablation)MLP terminality classifier · Terminality classification, bound 10accuracy: 90% -> 94%Improved
increase training set further beyond 10M (ablation)MLP terminality classifier · Terminality classification, bound 10accuracyNo clear effect
use classifier-assisted generation instead of conventional exact algorithmData generation of 100M varieties · Dimension 8 Picard rank 2 toric Fanocompute time: 160 days (600 CPU years) -> under 120 CPU hoursImproved

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