Research map / Learning theory and algorithms / Prediction and decision losses
Target Curricula via Selection of Minimum Feature Sets: a Case Study in Boolean Networks
Shannon Fenn, Pablo Moscato
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 |
|---|---|---|---|
| replace L1 loss with strict greatest-hierarchy loss L_gh | FBN with NAND gates · 7-bit cascaded parity, 6th target | test accuracy: 58% -> 84% | Improved |
| use L_gh instead of L1 | FBN with NAND gates · 7-bit cascaded parity | test accuracy: 16% reduction in training examples to exceed 90% mean test accuracy | Improved |
| order targets by easy-to-hard strictness (L_w -> L_lh -> L_gh) (ablation) | FBN target curriculum · five test-bed problems | test accuracy | Improved |
| use automatically detected target order instead of known order | L_gh · test-bed problems except CMUX | test accuracy | No clear effect |
| use automatically detected target order | L_gh · CMUX | test accuracy | Mixed |
| reverse the target order (tau = -1) (ablation) | L_gh · 5-bit addition | test accuracy | Hurt |
| use exact easy-to-hard order (tau = 1) over random orders (ablation) | L_gh · 5-bit addition | test accuracy | Improved |
| use random target order (ablation) | L_gh · 5-bit addition | test accuracy | No clear effect |
| add hierarchical losses with detected order | L_w/L_lh/L_gh · mammalian cell-cycle model | test accuracy | Mixed |
| use L_lh instead of L1 | FBN · E. coli target G8 | mean test accuracy difference vs L1: -0.9 | Hurt |
| use L_gh instead of L1 on later target in hierarchy | FBN · E. coli target G6 | mean test accuracy difference vs L1: 1.4 | Improved |
| use L_gh instead of L1 | FBN · Yeast SK -> {Ste9, Rum1} hierarchy | mean test accuracy difference vs L1: 0.9, 0.5, 0.3 | No clear effect |
Closest papers by meaning
- Sharp feature-learning transitions and Bayes-optimal neural scaling laws in extensive-width networks (NeurIPS 2026 · similarity 0.85)
- TRIM: Hybrid Inference via Targeted Stepwise Routing in Multi-Step Reasoning Tasks (ICLR 2026 · similarity 0.84)
- LightHGNN: Distilling Hypergraph Neural Networks into MLPs for 100x Faster Inference (ICLR 2024 · similarity 0.84)
- Finding the Minimal Parameter Budget for Implicit Reasoning: A Data Complexity Driven Scaling Law for Language Models (ICML 2026 · similarity 0.84)
- On the creation of narrow AI: hierarchy and nonlocality of neural network skills (NeurIPS 2025 · similarity 0.84)
- Verification Learning: Make Unsupervised Neuro-Symbolic System Feasible (ICML 2025 · similarity 0.83)
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.
