Research map / Probabilistic and scientific modeling / Molecules and proteins
Neural Low-Discrepancy Sequences
Michael Etienne Van Huffel, Nathan Kirk, Makram Chahine, Daniela Rus, T. Konstantin Rusch
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add discrepancy fine-tuning on top of Sobol' pretraining (vs Sobol') | NeuroLDS · D2 sym, d=4, N=10000 | discrepancy: 0.000167 -> 0.000098 | Improved |
| add NeuroLDS vs Sobol' | NeuroLDS · D2 star, d=4, N=100 | discrepancy: 0.009477 -> 0.008603 | Improved |
| add NeuroLDS vs Sobol' | NeuroLDS · D2 ctr, d=4, N=10000 | discrepancy: 0.000167 -> 0.000114 | Improved |
| add Sobol' pre-training before discrepancy fine-tuning (ablation) | NeuroLDS · 2d, 3d, 4d | point set quality | Improved |
| replace index-conditioned MLP with autoregressive GNN (ablation) | NeuroLDS · 2d | discrepancy | Hurt |
| increase sinusoidal bandwidth K from 8 to 32 (ablation) | NeuroLDS · 4d | stability of D2 sym curve | Improved |
| replace deep MLP with purely linear map (ablation) | NeuroLDS · 2d and 3d | discrepancy | Hurt |
| use one hidden layer instead of deeper MLP (ablation) | NeuroLDS · 2d and 3d | discrepancy | No clear effect |
| use length-proportional weights instead of uniform prefix weights (ablation) | NeuroLDS · d=4, N=10000 | prefix discrepancy | Mixed |
| use NeuroLDS points for training instead of Sobol' | MLP on Black-Scholes · 2D Black-Scholes | MSE (x1e-4): 4.04 -> 3.34 | Improved |
| use NeuroLDS ext discrepancy instead of Sobol' (ablation) | MLP on Black-Scholes · 2D Black-Scholes | MSE (x1e-4): 4.04 -> 4.01 | No clear effect |
| use NMGreedy instead of Sobol' points | MLP on Black-Scholes · 2D Black-Scholes | MSE (x1e-4): 4.04 -> 4.72 | Hurt |
Closest papers by meaning
- Inferring stochastic low-rank recurrent neural networks from neural data (NeurIPS 2024 · similarity 0.84)
- Universality of Linear Recurrences Followed by Non-linear Projections: Finite-Width Guarantees and Benefits of Complex Eigenvalues (ICML 2024 · similarity 0.83)
- No Wrong Turns: The Simple Geometry Of Neural Networks Optimization Paths (ICML 2024 · similarity 0.83)
- Efficient Discrepancy Testing for Learning with Distribution Shift (NeurIPS 2024 · similarity 0.83)
- Towards Understanding the Universality of Transformers for Next-Token Prediction (ICLR 2025 · similarity 0.83)
- Near-optimal Sketchy Natural Gradients for Physics-Informed Neural Networks (ICML 2025 · similarity 0.82)
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.
