1. 03 Oct, 2019 1 commit
  2. 06 Sep, 2019 1 commit
  3. 19 Apr, 2019 4 commits
  4. 11 Jan, 2019 1 commit
  5. 20 Oct, 2018 1 commit
  6. 15 Sep, 2018 1 commit
  7. 14 Sep, 2018 1 commit
  8. 17 Aug, 2018 1 commit
  9. 16 Aug, 2018 1 commit
    • Edi Prifti's avatar
      - add warning option · 6ff8aced
      Edi Prifti authored
      - improve documentation
      - improve digest plots => ggplot
      - finishing the vignette
      6ff8aced
  10. 15 Aug, 2018 1 commit
  11. 14 Aug, 2018 3 commits
  12. 13 Aug, 2018 1 commit
    • Edi Prifti's avatar
      - adding a new licence and updating description · 277a2141
      Edi Prifti authored
      - adding pos_score and neg_score to the model for exploration of the balance
      - add feature importance on prevalence PDA
      - fix order of visu in the mergeResults
      - adding custom colors in mergeMeltImportanceCV
      - adding plotModelScore function
      - adding vignette V0
      277a2141
  13. 16 Jul, 2018 1 commit
  14. 11 Jul, 2018 1 commit
    • Edi Prifti's avatar
      - updating version of the package · 9d42a5cb
      Edi Prifti authored
      - improving regression analyses, adding more attributes such as R2 and SER (standard error of the regression). default eval_to_fit is rsq_
      - fixing different bugs related to regression analyses as the code had substantialy evolved since this was tested
      - adding evalToFit attribute to models
      - improving overall code quality and robustness
      - controling color and shape in comparative plots
      - setting same objective in metal generators and unificators
      - adding family for glmnet in terda for regression mode
      9d42a5cb
  15. 31 May, 2018 1 commit
    • Edi Prifti's avatar
      - unify getBestModel methods in one · 26dfdd13
      Edi Prifti authored
      - don't evaluate ter models without two sides
      - increase robustness of the code
      - adding more isObject methods
      - updating version
      - adding printy and printObject methods
      - cleaning unused code
      - adding a modular testing markdown
      26dfdd13
  16. 22 Apr, 2018 1 commit
    • Edi Prifti's avatar
      - making checkers more robust · 9cc43acf
      Edi Prifti authored
      - fixing bugs in getFeaturePrevalence
      - adding penalty in selectBestPopulation
      - fixing bugs in mergeBestScoreCV
      - adding feature selection on mergeMeltImportanceCV
      - fixing small issues in plotComparativeEmpricalScore, plotComparativeCV & plotAbundanceByClass
      - adding importance in plotModel
      9cc43acf
  17. 30 Mar, 2018 1 commit
  18. 19 Mar, 2018 1 commit
  19. 10 Mar, 2018 1 commit
    • Edi Prifti's avatar
      - adding two functions analyzeImportanceFeatures and populationSetX · c0c62bb0
      Edi Prifti authored
      - fixing evaluation for k=1 for ratio
      - models in ratio without both terms are deleted
      - fixing bug in terbeam when no previous population is found
      - fixing feature importance, computing only for models for which the feature exists and only one seed
      - modifying printModel for ratio , better now
      - fixing bug in terga1 with printModel
      c0c62bb0
  20. 07 Mar, 2018 1 commit
  21. 06 Mar, 2018 1 commit
    • Edi Prifti's avatar
      - adding new functions such as isLearnerSota, computeCardEnrichment,... · 0fa97890
      Edi Prifti authored
      - adding new functions such as isLearnerSota, computeCardEnrichment, evaluateFeatureImportanceInPopulation and mergeMeltImportanceCV
      - fix mode train for AUC
      - fix bug in plotAbundanceByClass for population of 1 model
      - add mode compute.importance in fit
      - fix individual() function and its calls
      0fa97890
  22. 21 Feb, 2018 2 commits
    • Edi Prifti's avatar
      - making sure that accuracy is the only function that can change the signs. · ed4bab69
      Edi Prifti authored
      - computeInterect posClass is now levels(y)[1]
      - adding the class in printModel
      - improving network layout
      ed4bab69
    • Edi Prifti's avatar
      - adding analyzePopulationFeatures · 6fac5e33
      Edi Prifti authored
      - adding makeFeatureAnnot
      - adding makeFeaturePrevalenceNetworkCooccur
      - adding makeFeaturePrevalenceNetworkMiic
      - changing parameter order for mod in many functions
      - reorgnazing evaluation process
      - adding test mode for confusion matrix
      - adding isModelSotaGLMNET
      - modifying evaluation of AUC and accuracy for sota, outside of the way it is computed for BTR
      - adding updateModelIndex and updatePopulationIndex needed when we reorder X and restrict
      - improving filterFeatuesByPrevalence to have it work for more than 2 classes
      - fixing bug in digest
      adding max.nb.features in BTR and sota learners
      - improving terbeam ratio when many negative features are correlated strongly with y
      6fac5e33
  23. 30 Jan, 2018 1 commit
  24. 22 Jan, 2018 2 commits
    • Edi Prifti's avatar
      - terbeam using a max.nb.freature option so that for k=1 we don't build all... · a159b705
      Edi Prifti authored
      - terbeam using a max.nb.freature option so that for k=1 we don't build all the models, but rather use the clf$feature.cor object which is fast to select and make models
      - generateAllSingleFeatureModel focused in a subset
      - adding max.nb.features in terda
      - modifying core terga2 so that best models are kept for each sparsity rather than a whole population. This allows to solve many issues with when comparing results. Most of the best models will be in higher sparsities and lower ones won't be selected and lost. This ensures to keep a forced genetic variability for each sparsity
      - terga2 improve speed
      - fix getSign which gave inversed coefficients
      a159b705
    • Edi Prifti's avatar
      - adding BioQC package for speedup in wilcoxono · c9737245
      Edi Prifti authored
      - removing Hmisc, base cor.test used instead
      - fixing bug with compute_score (difference between matrix and data frame)
      - redoding getSign form more simplicty and efficiency
      - recoding filterfeaturesK for more speedup (testing with microbenchmark)
      - improving logging of the execution
      - adding printing in verbose in sota svm et rf
      c9737245
  25. 17 Jan, 2018 1 commit
    • Edi Prifti's avatar
      - adding Hmisc for rcorr of the filterFeatureK and updating version · ad709e66
      Edi Prifti authored
      - some stop assertions in evaluateYhat transofm in null returners
      - converting X to matrix in some functions which had unstable behaviour
      - roformating
      - adding the possibility to return statistical informaiton from functions on prevalence and differntial abundance plotPrevalence() and plotAbundanceByClass()
      ad709e66
  26. 12 Jan, 2018 1 commit
    • Edi Prifti's avatar
      - remove the Matrix dependency · 1abefb13
      Edi Prifti authored
      - remove calls to Matrix and make maj.table available for older versions
      - add feature cor to clf to improve speed in big datasets
      1abefb13
  27. 09 Jan, 2018 1 commit
    • Edi Prifti's avatar
      - adding library Matrix for large sparse matrices · 98b62edc
      Edi Prifti authored
      - adding X.max and X.min as an object in clf$data$ to avoid max(X) when infinite values in scores and min(X)
      - fixinb bug in digest when results without crossfold
      - improving generateAllSingleFeatureModel so that it is faster and works for big matrixes
      - adding evolve.k1 in terga2
      - improving speed in evolve2m
      98b62edc
  28. 05 Jan, 2018 1 commit
    • Edi Prifti's avatar
      - fix bugs on stops and adding try() callers · a1934ace
      Edi Prifti authored
      - fixing cleanPopulation tests
      - scoreRatio improving the sanity check
      - fixing some issues with getBestIndividual
      - fixing populationGet_X function and cases as well as its calls around the package
      - listOfModelsToModelCollection fixing issues and making more robust
      - fixing digest function as well as digestModelCollection
      - fixing issues with some plotComp functions as result to changes in digest
      - fixing some issues with metal main algorithm
      - fixing an issue with terbeam minimum number of individuals
      - adding some debuging for terda and adding capability to add new individuals for missing sparsities, based on features for close models
      - improving robustness of core terga1 engine
      - improving the evolve function in terga1
      a1934ace
  29. 08 Dec, 2017 1 commit
  30. 06 Dec, 2017 1 commit
    • Edi Prifti's avatar
      - adding documentaion on terga1 core functions · 2f80b2de
      Edi Prifti authored
      - fixing several bugs in terga1 ratio. When a model is not valid it is discarded and the order of the best samples changes. This is fixed now
      - another bug on listOfModelsToListOfSparseVec is fixed (by default it returns a vector)
      - improving the data filtering functions
      - fixing seed for terga1 core functions
      - making best_ancestor seeding more robust in the population function
      2f80b2de
  31. 21 Nov, 2017 1 commit
    • Edi Prifti's avatar
      - changing version · 91622af4
      Edi Prifti authored
      - exporting some functions
      - adding multiple time k-folding
      - increasing number of max compared experiments
      - adding metal per language
      91622af4
  32. 06 Oct, 2017 1 commit
  33. 21 Aug, 2017 1 commit