nabla
nabla dropped its C++ engine to chase exact derivatives at any order.
Comparison · Analytics
compositional.mle vs hubEvals
A side-by-side editorial comparison of compositional.mle and hubEvals — release velocity, themes, recent moves, and the top alternatives to consider.
Shared themes:r-package
compositional.mle vs hubEvals: at a glance
What is compositional.mle?
An MLE package rebuilt around composable solvers, then renamed to match.
compositional.mle performs numerical maximum likelihood estimation in R, with the optimisation strategy expressed as composed pieces rather than configured up front. It began in November 2025 as numerical.mle, a configuration-object package with fixed solvers. The v0.2.0 rewrite replaced that with solver factories sharing a uniform signature and operators for chaining and racing them, and renamed the package accordingly. The two most recent releases are CRAN submission work.
What is hubEvals?
Forecast-hub scoring that learned to handle joint, sample-based predictions.
hubEvals scores model output from collaborative forecasting hubs, wrapping scoringutils and translating hubverse formats into forecast objects it can evaluate. The package has moved quickly from a thin translation layer to something that handles every output type the hubverse defines — quantile, mean, median, nominal and ordinal pmf, and samples. The most recent releases are almost entirely about the failure modes of relative skill scoring rather than about new metrics.
compositional.mle vs hubEvals: editorial side-by-side
compositional.mle
0.0ANALYTICS
An MLE package rebuilt around composable solvers, then renamed to match.
◆ Current state
compositional.mle performs numerical maximum likelihood estimation in R, with the optimisation strategy expressed as composed pieces rather than configured up front. It began in November 2025 as numerical.mle, a configuration-object package with fixed solvers. The v0.2.0 rewrite replaced that with solver factories sharing a uniform signature and operators for chaining and racing them, and renamed the package accordingly. The two most recent releases are CRAN submission work.
◆ Where it's heading
The arc is a design idea overtaking an implementation: version 0.1.0 exposed configuration functions and named solvers, version 0.2.0 turned solvers into values that can be sequenced with %>>%, raced with %|%, restarted, or conditionally refined, and separated the statistical problem from the optimisation strategy. Since then all effort has gone into CRAN acceptance, dead code removal, policy compliance, validation fixes. That is a package that redesigned itself early and is now trying to get through the door.
◆ Prediction
With the composable API settled, the next work will most likely be additional solvers and transformers plugged into the existing operators rather than another redesign.
hubEvals
2.5ANALYTICS
Forecast-hub scoring that learned to handle joint, sample-based predictions.
◆ Current state
hubEvals scores model output from collaborative forecasting hubs, wrapping scoringutils and translating hubverse formats into forecast objects it can evaluate. The package has moved quickly from a thin translation layer to something that handles every output type the hubverse defines — quantile, mean, median, nominal and ordinal pmf, and samples. The most recent releases are almost entirely about the failure modes of relative skill scoring rather than about new metrics.
◆ Where it's heading
Two threads dominate. The first is coverage of output types, which reached its widest point with sample-based and compound scoring. The second, and the one occupying every recent release, is making relative skill degrade gracefully: single-model input, comparison groups with one model, and groups missing the requested baseline have each been converted from a cryptic upstream abort into a defined result. That pattern — inherited scoringutils errors being caught and given hub-specific meaning — is the clearest signal of where this package adds value.
◆ Prediction
Expect continued work smoothing scoringutils error surfaces into hub-aware behaviour, and performance attention on relative skill, which was explicitly optimised in the latest release.
Alternatives to compositional.mle and hubEvals
Other Analytics products tracked by Sparkpulse, ranked by recent ship velocity. Each card links to a full editorial trajectory and lets you pivot into a head-to-head comparison with either compositional.mle or hubEvals.
snowflakeauth
Eight months from first release to keyring caching and workload identity.
kwb.geosalz
A research-project workflow package where the interesting work is in the plumbing.
cyclocomp
A cyclomatic complexity checker that ships once every couple of years, and lands when it does.
revdbayes
Extreme value sampling in pure upkeep mode, mostly answering to Rcpp and CRAN.
lazyeval
A package retired in 2017 just got rewritten against R's public C API.
See all compositional.mle alternatives → · See all hubEvals alternatives →
Recent activity from compositional.mle and hubEvals
Latest ship moves from both products, interleaved chronologically. ⚡ = editorial spark.
- 24d agohubEvalsScored-forecast counts and faster relative skill
- 1mo agohubEvalsDisaggregated relative skill no longer aborts the whole call
- 1mo agohubEvalsSingle-model scoring returns relative skill of 1 instead of erroring
- 5mo agohubEvalsSample output types and multivariate compound scoring ⚡
- 6mo agocompositional.mleParallel racing fixed under the future package
- 6mo agocompositional.mleDead code removed and CRAN policy compliance work
- 6mo agohubEvalsScoring on transformed scales via transform arguments
- 8mo agocompositional.mleSolvers become composable values, and the package is renamed ⚡
- 8mo agocompositional.mleFirst release as numerical.mle, built on configuration objects
- 11mo agohubEvalsFirst release: score_model_out() and the scoringutils bridge
Frequently asked questions
What is the difference between compositional.mle and hubEvals?
Both compete on the same themes — r-package — within Analytics. hubEvals is currently shipping more aggressively (velocity 2.5 vs 0.0), with 0 editorial sparks in the last 30 days against 0. See the at-a-glance table above for a side-by-side breakdown of velocity, recent sparks, and editorial themes.
Is compositional.mle better than hubEvals?
Sparkpulse doesn't pick a winner — we score release velocity, not feature parity. hubEvals is currently shipping more aggressively (velocity 2.5 vs 0.0), with 0 editorial sparks in the last 30 days against 0. For your specific use case, the alternatives sections above list other Analytics products to evaluate alongside.
What are the best alternatives to compositional.mle?
Top compositional.mle alternatives in Analytics are ranked by recent ship velocity. Browse the "compositional.mle alternatives" section above for the current picks, or visit /alternatives/compositional-mle for the full list with editorial commentary on each.
What are the best alternatives to hubEvals?
Top hubEvals alternatives in Analytics are ranked by recent ship velocity. Browse the "hubEvals alternatives" section above for the current picks, or visit /alternatives/hubevals for the full list with editorial commentary on each.