A Proper Scoring Rule for Virtual Staining
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866918358565257216 |
|---|---|
| author | Tonks, Samuel Hood, Steve Musso, Ryan Hopely, Ceridwen Titus, Steve Doan, Minh Styles, Iain Krull, Alexander |
| author_facet | Tonks, Samuel Hood, Steve Musso, Ryan Hopely, Ceridwen Titus, Steve Doan, Minh Styles, Iain Krull, Alexander |
| contents | Generative virtual staining (VS) models for high-throughput screening (HTS) can provide an estimated posterior distribution of possible biological feature values for each input and cell. However, when evaluating a VS model, the true posterior is unavailable. Existing evaluation protocols only check the accuracy of the marginal distribution over the dataset rather than the predicted posteriors. We introduce information gain (IG) as a cell-wise evaluation framework that enables direct assessment of predicted posteriors. IG is a strictly proper scoring rule and comes with a sound theoretical motivation allowing for interpretability, and for comparing results across models and features. We evaluate diffusion- and GAN-based models on an extensive HTS dataset using IG and other metrics and show that IG can reveal substantial performance differences other metrics cannot. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_23305 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | A Proper Scoring Rule for Virtual Staining Tonks, Samuel Hood, Steve Musso, Ryan Hopely, Ceridwen Titus, Steve Doan, Minh Styles, Iain Krull, Alexander Machine Learning Generative virtual staining (VS) models for high-throughput screening (HTS) can provide an estimated posterior distribution of possible biological feature values for each input and cell. However, when evaluating a VS model, the true posterior is unavailable. Existing evaluation protocols only check the accuracy of the marginal distribution over the dataset rather than the predicted posteriors. We introduce information gain (IG) as a cell-wise evaluation framework that enables direct assessment of predicted posteriors. IG is a strictly proper scoring rule and comes with a sound theoretical motivation allowing for interpretability, and for comparing results across models and features. We evaluate diffusion- and GAN-based models on an extensive HTS dataset using IG and other metrics and show that IG can reveal substantial performance differences other metrics cannot. |
| title | A Proper Scoring Rule for Virtual Staining |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2602.23305 |