Research map / Label-efficient and robust learning / Recognition and long-tail
Combining Logical Abduction and Statistical Induction: Discovering Written Primitives with Human Knowledge
Wang-Zhou Dai, Zhi-Hua Zhou
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 FOL stroke knowledge via abduction before sparse coding | Sparse Coding dictionary · MNIST classification, |D|=200 | accuracy (%): 95.77 -> 97.01 | Improved |
| add FOL stroke knowledge | Sparse Coding dictionary · Omniglot small 1 classification | accuracy (%): |D|=20: 16.46 -> 16.24; |D|=200: 25.40 -> 26.64 | Mixed |
| add FOL stroke knowledge, domain adaptation MNIST to Omniglot | Sparse Coding dictionary · M2O adaptation, |D|=200 | accuracy (%): 18.62 -> 24.63 | Improved |
| add FOL stroke knowledge, domain adaptation Omniglot to MNIST | Sparse Coding dictionary · O2M adaptation, |D|=200 | accuracy (%): 96.11 -> 96.39 | No clear effect |
| replace stroke knowledge with 2-cluster k-means knowledge (ablation) | LASIN · Omniglot small 1, |D|=100 | accuracy (%): 24.74 -> 23.26 | Hurt |
| replace stroke knowledge with 2-cluster k-means knowledge (ablation) | LASIN · MNIST classification, |D|=200 | accuracy (%): 97.01 -> 96.91 | No clear effect |
| replace stroke knowledge with 2-cluster knowledge (ablation) | LASIN · M2O adaptation, |D|=200 | accuracy (%): 24.63 -> 22.57 | Hurt |
| LASIN versus patch-based sparse coding with larger code length | MNIST classification · MNIST | accuracy (%): LASIN |D|=100 96.27 vs patched SC 100x4 96.41 | Mixed |
Closest papers by meaning
- Ambiguity-Aware Abductive Learning (ICML 2024 · similarity 0.83)
- Phenomenal Yet Puzzling: Testing Inductive Reasoning Capabilities of Language Models with Hypothesis Refinement (ICLR 2024 · similarity 0.83)
- From Noise to Diversity: Random Embedding Injection in LLM Reasoning (NeurIPS 2026 · similarity 0.83)
- Curriculum Abductive Learning (NeurIPS 2025 · similarity 0.82)
- On the Inductive Bias of Stacking Towards Improving Reasoning (NeurIPS 2024 · similarity 0.82)
- Safe Abductive Learning in the Presence of Inaccurate Rules (AAAI 2024 · 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.
