v2: Allow infinite `nominalValue`
還沒有人認領這個 Issue。
評估
研究方向
Start by reading the PEtab v2 data format documentation and development requirements linked in the issue, then inspect how the specification defines and validates nominalValue. Compare the proposed allowance for inf and -inf with the existing handling of condition table values and the related issues #682 and #684. Done means the specification clearly permits infinite nominal values, keeps NaN invalid, and explains the behavior for estimated and fixed parameters.
由索引模型根據 Issue 內容生成。
描述
In PEtab 2.0.0, nominalValue is NUMERIC, i.e., finite. The resolution of #682 kept it that way. We propose to allow inf and -inf as well, as erratum E2 to PEtab 2.0.0.
Motivation
PEtab v2 removed parameterScale. The list of changes from v1 says: "To obtain the same effect, the model parameters can be transformed in the model file" (documentation_data_format.rst#L176-L182). The development docs require that problems from earlier versions remain specifiable (development.rst#L249-L251).
With finite nominal values, a transformed parameter doesn't give the same effect. In v1, a parameter k1 with parameterScale=log can be fixed at nominalValue=0 (v1 nominal values are on linear scale). This is a common way to switch off a mechanism without changing the model. After the recommended transformation to log_k1, the same fixed value is -inf, which NUMERIC doesn't allow.
Proposal
nominalValuebecomesNUMERIC or 'inf' or '-inf'.NaNstays invalid.- For
estimate=false, the value is applied as specified. Undefined results such as0 * infare the responsibility of the model author. This already applies to infinite values set via the condition table, whose math expressions supportinfand-inf. - For
estimate=true, the nominal value has no role in the problem definition, so allowing infinite values there has no consequences. Among other things, it allows storing estimation results at the boundary of an infinite parameter domain.
Why an erratum
Like #682, this resolves a conflict between the stated intent of the spec (parameterScale can be replaced by transforming the model) and a type restriction. Admittedly, the conflicting statement is in the informative list of changes, and converters can avoid the issue by not transforming fixed parameters. Still, the restriction doesn't serve a purpose, and allowing infinite values makes the statement about replacing parameterScale accurate.
libpetab-python currently doesn't reject infinite nominal values, so existing tooling already accepts them.
Relates to #682, #684.
🤖 Generated with Claude Code
- 主要語言
- 沒有語言資料
- 星號
- 67
- 分支
- 16
- 平均合併
- 1 天 19 小時
- 30 天內合併 PR
- 1
環境準備
- 沒有 Dockerfile 或 Docker Compose 檔案
- 沒有 Pull Request 範本
- 閱讀貢獻指南
從這裡開始
- 先讀完整個 Issue，再讀專案的貢獻指南。
- 在 Issue 下留言說明你要接手 —— 這能避免兩個人做同樣的事。
- Fork 儲存庫，在一個分支上完成修改。
- 送出 Pull Request，並在描述裡引用這個 Issue 編號。
PEtab-dev/PEtab 的其他 Issue
-
v2
難度 4/5 1-2 天 新手友好度 20/100
-
Docs: v1→v2 migration guide (incl. parameter scaling) and comparing objective values across tools可能已有人在做 @dweindl 於 10 天前認領。 未關閉documentation
-
難度 5/5 一週以上 新手友好度 25/100
-
Poisson and negative binomial distributions for parameter priors and noise models可能已有人在做 @dweindl 於 325 天前認領。 未關閉v2
-
v2: Add how-tos未關閉documentation
難度 5/5 一週以上 新手友好度 25/100
相似的 Issue
-
neuromuscular obsolete
難度 2/5 1-3 小時 新手友好度 68/100
monarch-initiative/mondo#10816 ·
維護者通常 2 天內回覆
-
correction metadata
難度 2/5 1-3 小時 新手友好度 76/100
acl-org/acl-anthology#10558 ·
維護者通常 1 天內回覆
-
Type: ☹︎ Bug
難度 2/5 1-3 小時 新手友好度 78/100
posit-dev/great-tables#1028 ·
維護者通常 1 天內回覆
-
content good first issue
難度 2/5 1-3 小時 新手友好度 78/100
-
難度 2/5 1-3 小時 新手友好度 78/100