Research map / Probabilistic and scientific modeling / Molecules and proteins
Human spatiotemporal pattern learning as probabilistic program synthesis
Tracey Mills, Joshua B. Tenenbaum, Samuel J Cheyette
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 Bayesian Ridge regression baseline with LoT program-synthesis model | LoT vs Ridge · Human 2D sequence prediction | correlation with human per-sequence error: 0.23 -> 0.56 | Improved |
| replace GP-NC with compositional GP-SL | GP model · Human 2D sequence prediction | correlation with human per-sequence error: 0.48 -> 0.50 | Improved |
| replace GP-SL with structured LoT model | GP-SL · Human 2D sequence prediction | correlation with human per-sequence error: 0.50 -> 0.56 | Improved |
| replace GP-NC with LoT model | GP-NC · Human 2D sequence prediction | delta AIC / log-likelihood: -31592 -> -21835 | Improved |
| replace GP-SL with LoT model | GP-SL · Human 2D sequence prediction | log-likelihood of human data: -28416 -> -21835 | Improved |
| replace Bayesian Ridge with GP-SL (compositional kernels) | Ridge · Human 2D sequence prediction | log-likelihood of human data: -30187 -> -28416 | Improved |
| replace GP-NC with Ridge regression | GP-NC · Human 2D sequence prediction | log-likelihood of human data: -31592 -> -30187 | Improved |
| add structure to the hypothesis space (GP-NC to GP-SL) | GP-NC · Human 2D sequence prediction | delta AIC: 19514 -> 13162 | Improved |
| more observed points (first to final timepoint) | human participants · Human 2D sequence prediction | mean relative error: 0.94 -> 0.21 | Improved |
| more observed points beyond timepoint 10 | GP-NC and Ridge models · Human 2D sequence prediction | relative error over time | No clear effect |
| more observed points | LoT and GP-SL models · Human 2D sequence prediction | relative error over time | Improved |
| sequence complexity rating (higher) | human participants · Human 2D sequence prediction | mean relative error: r = 0.77 | Hurt |
Closest papers by meaning
- In-Context Learning Dynamics with Random Binary Sequences (ICLR 2024 · similarity 0.84)
- Emergence of Hidden Capabilities: Exploring Learning Dynamics in Concept Space (NeurIPS 2024 · similarity 0.84)
- In-Context Learning Strategies Emerge Rationally (NeurIPS 2025 · similarity 0.84)
- Understanding the Learning Dynamics of Alignment with Human Feedback (ICML 2024 · similarity 0.84)
- Turning large language models into cognitive models (ICLR 2024 · similarity 0.83)
- Synthesizing Images on Perceptual Boundaries of ANNs for Uncovering and Manipulating Human Perceptual Variability (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.
