Research map / Probabilistic and scientific modeling / Variational and generative sampling
Characteristic Circuits
Zhongjie Yu, Martin Trapp, Kristian Kersting
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 |
|---|---|---|---|
| Use RDC-based structure learning instead of G-test-based structure learning (ablation) | Characteristic circuits with α-stable leaves · Heterogeneous UCI datasets | test log-likelihood | Mixed |
| Use α-stable rather than normal continuous leaves (ablation) | Characteristic circuits · Heterogeneous UCI datasets | test log-likelihood | Improved |
| Model discrete variables as normal leaves rather than categorical leaves (ablation) | CC-N · MM synthetic data | CFD: 0.0270 vs 0.0006 | Hurt |
| Learn parameters by minimizing CFD (ablation) | Random-structure characteristic circuits · MM synthetic data | test log-likelihood: -4.93 -> -3.50 | Improved |
| Learn parameters by minimizing CFD (ablation) | Random-structure characteristic circuits · BN synthetic data | test log-likelihood: -6.30 -> -4.12 | Improved |
| Learn the circuit structure from data instead of using a random structure, then learn parameters (ablation) | Characteristic circuits · MM synthetic data | test log-likelihood: -3.50 -> -2.86 | Improved |
| Learn the circuit structure from data instead of using a random structure, then learn parameters (ablation) | Characteristic circuits · BN synthetic data | test log-likelihood: -4.12 -> -3.27 | Improved |
| Use randomized weights and leaf parameters rather than structure-learning initialization (ablation) | Structure-learned characteristic circuits with parameter learning · MM synthetic data | test log-likelihood: -2.86 -> -3.34 | Hurt |
| Use randomized weights and leaf parameters rather than structure-learning initialization (ablation) | Structure-learned characteristic circuits with parameter learning · BN synthetic data | test log-likelihood: -3.27 -> -3.93 | Hurt |
| Use a structured characteristic circuit instead of a single empirical characteristic function | Characteristic circuit with ECF leaves · MM and BN synthetic data | CFD | Improved |
| Use RDC-based rather than G-test-based product-node splitting (ablation) | Structure-learned characteristic circuits · Heterogeneous UCI datasets | test log-likelihood | Improved |
Closest papers by meaning
- On the Hardness of Approximating Distributions with Tractable Probabilistic Models (NeurIPS 2025 · similarity 0.89)
- Scaling Probabilistic Circuits via Monarch Matrices (ICML 2025 · similarity 0.89)
- On the Expressive Power of Tree-Structured Probabilistic Circuits (NeurIPS 2024 · similarity 0.88)
- The Expressivity Boundary of Probabilistic Circuits: A Comparison with Large Language Models (NeurIPS 2026 · similarity 0.86)
- Probabilistic Neural Circuits (AAAI 2024 · similarity 0.86)
- A Unified Framework for Human-Allied Learning of Probabilistic Circuits (AAAI 2025 · similarity 0.86)
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.
