Research map / Large language models / AI and society
Cross-Corpora Unsupervised Learning of Trajectories in Autism Spectrum Disorders
Huseyin Melih Elibol, Vincent Nguyen, Scott Linderman, Matthew Johnson, Amna Hashmi, Finale Doshi-Velez
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 dynamic (stick-breaking DTM) topics instead of static LDA | LDA · ASD EHR and SM held-out words | held-out predictive log-likelihood | Improved |
| increase LDA topics to 15 (match number of yearly topics) | LDA · ASD EHR and SM held-out words | held-out predictive log-likelihood | Hurt |
| share one topic-proportion vector per patient (theta_p) across their notes (ablation) | SB-DTM-θd · ASD EHR and SM held-out words | held-out predictive log-likelihood | Improved |
| add cross-corpora structure (shared dynamics, corpus-specific terms) (ablation) | SB-DTM-θp · combined EHR+SM held-out words | held-out predictive log-likelihood | Improved |
| assume constant per-patient topic proportions (DTM-θp) (ablation) | SB-DTM-θd · future notes at age seven and above (EHR, SM) | predictive likelihood of future notes | Improved |
| add cross-corpora transfer (ccDTM) on combined data (ablation) | SB-DTM-θp · future notes, combined EHR+SM | predictive likelihood of future notes | Improved |
| use DTM-θp instead of logistic regression for predicting future conditions | Logistic regression · future CUIs with at least 10% prevalence | AUC | Improved |
| use DTM-θp instead of copying past conditions forward | Copy past · future CUIs with at least 10% prevalence | AUC | Improved |
| train flat LDA on combined EHR+SM instead of EHR alone | LDA · EHR held-out data | predictive likelihood | Improved |
| train flat models on combined EHR+SM instead of SM alone | LDA · SM held-out data | predictive likelihood | Hurt |
| use cross-corpora DTM when training on combined data | SB-ccDTM · EHR and SM, random and future held-out | predictive likelihood | Improved |
| use cross-corpora DTM instead of per-patient DTM for speed | SB-DTM-θp · full corpus | seconds per iteration: 2.14 -> 0.953 | Improved |
Closest papers by meaning
- Unscrambling disease progression at scale: fast inference of event permutations with optimal transport (NeurIPS 2024 · similarity 0.84)
- Understanding Unique Behavioral Patterns through Multimodal Analysis of Eye-Hand Coordination in Autistic Children (Student Abstract) (AAAI 2025 · similarity 0.83)
- Risk Horizons: Structured Hypothesis Spaces for Longitudinal Clinical Prediction (NeurIPS 2026 · similarity 0.82)
- DECT: Harnessing LLM-assisted Fine-Grained Linguistic Knowledge and Label-Switched and Label-Preserved Data Generation for Diagnosis of Alzheimer’s Disease (AAAI 2025 · similarity 0.82)
- Multimodal Disease Progression Modeling via Spatiotemporal Disentanglement and Multiscale Alignment (NeurIPS 2025 · similarity 0.82)
- Exploiting Negative Samples: A Catalyst for Cohort Discovery in Healthcare Analytics (ICML 2024 · similarity 0.81)
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.
