fhollenbach / ebma Goto Github PK
View Code? Open in Web Editor NEWRepository for EBMA Package with Estimation via EM Algorithm or Gibbs Sampling
Repository for EBMA Package with Estimation via EM Algorithm or Gibbs Sampling
Possible improvement of coding gibbs samplers in stan instead. Would allow us to use stan infrastructure for multiple cores, convergence tests, etc.
EBMA/EBMAforecast/R/fitEnsembleNormal.R
Line 47 in 2961a45
The quoted code sum(W[1,]) == 1
suffers from numerical imprecision if the number of models is sufficiently large. On my Mac M3 this check fails on a structure with M=10 models. I suggest to replace the call with
all.equal(sum(W[1,]), 1)
, which is more tolerant to floating point imprecision.
I did not check it, but the problem might occur in other places.
P.S. Reproducible example (on Apple M3, might not occur on other machines):
>
>
> # Minimal reproducible example to show problem on Apple M3
> library(EBMAforecast)
> set.seed(1)
> pred_train <- matrix( rnorm(900), ncol=10)
> pred_test <- matrix( rnorm(100), ncol=10)
> truth_train <- apply(pred_train,1, mean) + rnorm(nrow(pred_train), sd=0.1)
> truth_test <- apply(pred_train,1, mean) + rnorm(nrow(pred_test), sd=0.1)
> myForecastData <- makeForecastData(
+ .predCalibration = pred_train,
+ .outcomeCalibration = truth_train,
+ .predTest = pred_test,
+ .outcomeTest = truth_test
+ )
> calibrateEnsemble(myForecastData, model="normal")
Model weights estimated using EM algorithmError in .local(.forecastData, tol, maxIter, method, exp, useModelParams, :
Vector of initial model weights must sum to 1.
> # Problem is in the comparison
> M <- dim(myForecastData@predCalibration)[2]
> W <- matrix(rep(1/M, M, collapse=FALSE), ncol=M)
> W
[,1] [,2] [,3] [,4] [,5] [,6] [,7] [,8] [,9] [,10]
[1,] 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1
> dim(W)[2] != M
[1] FALSE
> sum(W[1,]) == 1
[1] FALSE
> all.equal(sum(W[1,]), 1)
[1] TRUE
> sessionInfo()
R version 4.3.2 (2023-10-31)
Platform: aarch64-apple-darwin20 (64-bit)
Running under: macOS Sonoma 14.2
Matrix products: default
BLAS: /System/Library/Frameworks/Accelerate.framework/Versions/A/Frameworks/vecLib.framework/Versions/A/libBLAS.dylib
LAPACK: /Library/Frameworks/R.framework/Versions/4.3-arm64/Resources/lib/libRlapack.dylib; LAPACK version 3.11.0
locale:
[1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8
time zone: Europe/Berlin
tzcode source: internal
attached base packages:
[1] stats graphics grDevices utils datasets methods base
other attached packages:
[1] EBMAforecast_1.0.31
loaded via a namespace (and not attached):
[1] utf8_1.2.4 generics_0.1.3 gtools_3.9.5 separationplot_1.4
[5] stringi_1.8.3 digest_0.6.34 magrittr_2.0.3 RColorBrewer_1.1-3
[9] evaluate_0.23 grid_4.3.2 fastmap_1.1.1 plyr_1.8.9
[13] nnet_7.3-19 backports_1.4.1 Formula_1.2-5 gridExtra_2.3
[17] purrr_1.0.2 fansi_1.0.6 scales_1.3.0 abind_1.4-5
[21] cli_3.6.2 rlang_1.1.3 munsell_0.5.0 Hmisc_5.1-1
[25] base64enc_0.1-3 yaml_2.3.8 tools_4.3.2 checkmate_2.3.1
[29] htmlTable_2.4.2 dplyr_1.1.4 colorspace_2.1-0 ggplot2_3.5.0
[33] vctrs_0.6.5 R6_2.5.1 rpart_4.1.21 lifecycle_1.0.4
[37] stringr_1.5.1 htmlwidgets_1.6.4 MASS_7.3-60 foreign_0.8-85
[41] cluster_2.1.4 pkgconfig_2.0.3 pillar_1.9.0 gtable_0.3.4
[45] glue_1.7.0 data.table_1.15.0 Rcpp_1.0.12 xfun_0.42
[49] tibble_3.2.1 tidyselect_1.2.0 rstudioapi_0.15.0 knitr_1.45
[53] htmltools_0.5.7 rmarkdown_2.25 compiler_4.3.2
>
throw a warning, but allow the calibration of weights if test set is empty
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.