{"paper":{"title":"Bayesian Nonparametric Mixed-Effect ODEs with Gaussian Processes","license":"http://creativecommons.org/licenses/by/4.0/","headline":"A Bayesian nonparametric model decomposes each subject's ODE vector field into a shared population Gaussian process and subject-specific deviations.","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Harri L\\\"ahdesm\\\"aki, Julien Martinelli, Maksim Sinelnikov, M\\'elanie Prague, Quentin Clairon","submitted_at":"2026-05-13T06:57:01Z","abstract_excerpt":"Dynamical modelling is central to many scientific domains, including pharmacometrics, systems biology, physiology, and epidemiology. In these settings, heterogeneity is often intrinsic: different subjects or units follow related but distinct continuous-time dynamics. Classical nonlinear mixed-effects Ordinary Differential Equation (ODE) models address this by combining population-level structure with subject-specific effects, but they rely on a parametric vector field and are therefore vulnerable to structural misspecification and unmodelled mechanisms. This motivates nonparametric approaches "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Across controlled heterogeneous ODE benchmarks spanning oscillatory, biomedical systems, MEGPODE improves population-field recovery and subject-level trajectory prediction relative to strong baselines.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The vector field of each subject can be usefully decomposed into a shared population component and a subject-specific deviation, both endowed with Gaussian process priors, and that virtual collocation observations suffice to avoid repeated ODE solves while preserving accurate posterior inference.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"MEGPODE decomposes subject-specific ODE vector fields into population and individual Gaussian process priors and uses Kalman smoothing with virtual collocation to enable efficient Bayesian mixed-effects inference for heterogeneous dynamical systems.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A Bayesian nonparametric model decomposes each subject's ODE vector field into a shared population Gaussian process and subject-specific deviations.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"e4a05f2b72b83247f53e9d897a99bdceeb01c54db77515c67fa2e4bcbcb55f1e"},"source":{"id":"2605.13088","kind":"arxiv","version":1},"verdict":{"id":"aa688046-3a2e-4e27-9b6d-00bdf8b7b726","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T19:44:47.057371Z","strongest_claim":"Across controlled heterogeneous ODE benchmarks spanning oscillatory, biomedical systems, MEGPODE improves population-field recovery and subject-level trajectory prediction relative to strong baselines.","one_line_summary":"MEGPODE decomposes subject-specific ODE vector fields into population and individual Gaussian process priors and uses Kalman smoothing with virtual collocation to enable efficient Bayesian mixed-effects inference for heterogeneous dynamical systems.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The vector field of each subject can be usefully decomposed into a shared population component and a subject-specific deviation, both endowed with Gaussian process priors, and that virtual collocation observations suffice to avoid repeated ODE solves while preserving accurate posterior inference.","pith_extraction_headline":"A Bayesian nonparametric model decomposes each subject's ODE vector field into a shared population Gaussian process and subject-specific deviations."},"references":{"count":44,"sample":[{"doi":"","year":2023,"title":"Prediction of long- term humoral response induced by the two-dose heterologous ad26.zebov, mva-bn-filo vaccine against ebola.npj Vaccines, 2023","work_id":"1eea5dad-aa74-4352-9b1a-0c7e2da35deb","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"An amortized approach to non-linear mixed-effects modeling based on neural posterior estimation","work_id":"a7f99175-131b-4cc7-b98d-1b6c97358629","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2006,"title":"Christopher M Bishop and Nasser M Nasrabadi.Pattern recognition and machine learning. Springer, 2006","work_id":"5da02df6-468d-42ac-9ce5-e8b2662f9f82","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2004,"title":"Cambridge University Press","work_id":"27c44c9f-fc3a-4dda-8743-a5b72e58fafc","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"Dominic Stefan Bräm, Bernhard Steiert, Marc Pfister, Britta Steffens, and Gilbert Koch. Low- dimensional neural ordinary differential equations accounting for inter-individual variability implemented ","work_id":"cee96acb-8deb-4692-89cb-c6acad96cf2f","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":44,"snapshot_sha256":"bc9984c39009a2f452d9e728d2792b3906e1424b4854f68bdfe5ba5dfe73a328","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}