Edit on GitHub

New

Version 0.4.2

  • News

    • sampwsp: Build training vs. test sets by WSP sampling.

  • Modifications

    • Code cleaning.

Version 0.4.1

  • Breaking

    • savgk: order of arguments have changed.

  • Modifications

    • Code cleaning.

Version 0.4.0

Warning: Major breaking changes with versions 0.3. The embedded syntax has changed, with the use of the new function model: - 'mod = plskern(; nlv = 15)' is now writen as 'mod = model(plskern; nlv = 15)'.

  • Modifications
    • Code cleaning. The examples in th the function help pages have
    been corrected from typing errors.

Version 0.3.7

  • News

    • aggsum: Compute sub-total sums by class for a categorical variable.
    • findindex: Replace a vector containg levels by the indexes of a set of levels.
    • getknn: Add of angular and correlation distances.
    • kplslda, kplsqda, kplskdeda: Kernel PLS-LDA, PLS-QDA, PLS-KDEDA.
    • dkplslda, dkplsqda, dkplskdeda: Direct kernel PLS-LDA, PLS-QDA, PLS-KDEDA.
    • mbplslda, mbplsqda, mbplskdeda: Multiblock PLS-LDA, PLS-QDA, PLS-KDEDA.
    • merrp: Mean intra-class classification error rate.
    • mweightcla: Compute observation weights for a categorical variable, given specified sub-total weights for the classes.
    • rownorm: Row-wise norms.
    • snorm: Row-wise norming of X-data.

  • Breaking

    • confusion: renamed to conf, and output 'accuracy' renamed to 'accpct' and modified.

  • Modifications

    • FDA and predictive DA functions: improvements to better take into account for unbalanced classes.
    • Code cleaning.

Version 0.3.6

  • Modifications
    • Code improvement.

Version 0.3.5

  • Modifications
    • Code improvement.

Version 0.3.4

  • News

    • colmed: Column-wise medians.

  • Modifications

    • Code improvement.

Version 0.3.3

  • Modifications
    • Code improvement.
    • calds, calpds: Function 'fit!' was added.
  • Breaking
    • rmgap: works now with a function 'transf'. Function rmgap! was removed.

Version 0.3.2

  • Modifications

    • Code improvement.

  • Breaking

    • hconcat renamed to mbconcat
    • recodnum2cla renamed to recodnum2int
    • viperm! renamed to viperm

Version 0.3.1

  • New

    • hconcat Transformer contatenating multi-block X-data.
    • blockscal Scaling of multiblock X-data. Replace functions 'fblockscal_...".
    • mbplsrda Discrimination based on multiblock partial least squares regression (MBPLSR-DA).

  • Modifications

    • Code improvement.
    • mbplsr, soplsr : correction when 'scal' = true.

Version 0.3.0

Warning: Major breaking changes. Package Jchemo has been deeply restructured, to enable an "embedded" syntax and facilitate pipelines building. The previous "direct" syntax is still allowed (although arugments have changed for some functions, see the help pages) but is not favor anymore. Users who prefer the previous syntax will have to keep working with versions < 0.3.0. Sorry for any inconveniance.

Some typing errors may have been introduced due to the restructuration. They will be corrected in versions > 0.3.0.

Some specific modified points are:

  • In the arguments, all String types have been replaced by Type Symbol (e.g. "unif" is replaced by :unif)
  • The 'weights' (row weighting in some functions, such as pcasvd etc.) argument must now be of type 'Weight', built from function 'mweight'.
  • Sampling functions 'samp...' have changed.
  • Function 'mtest' renamed to 'sampdf'.
  • Functions center, scale, cscale, blockscale renamed to fcenter, fscale, etc. Alternatively, new transformers center, scale and cscale have been created.
  • Functions isel and viperm remated to isel! and viperm!, and syntax changed.
  • Syntax of tuning functions gridscore and gridcv has changed, and the functions are now genereic (no need anymore to call specific functions ...lv and ...lb).
  • Removed function (temporary or not):
    • baggr and its utilities
    • cplsravg
    • gridcvmb, gridcvlvmb
    • interpl_mon
    • lwmlrs, lwplsrs, lwmlrdas, lwlsrdas (since can be built from pipelines)
    • lwplsrdaavg, lwplsldaavg, lwplsqdaavg
    • mavg_runmean
    • mbunif, mbwcov
    • nsc, nscrda, nscda
    • occknndis, occlknndis
    • plsrdaavg, plsldaavg, plsqdaavg

Version 0.2.4

  • Modifications
    • Code improvement.

Version 0.2.3

  • New

    • samprand: Build training vs. test sets by random sampling.
    • svmr, svmda: SVM (wrappers to LIBSVM.jl) were reset in the package.

  • Modifications

    • baggr has been parallelized (multi-threading).
    • Code improvement.

  • Breaking changes

    • sampks, sampdp, samprand, sampsys, sampcla: order of arguments have changed.
    • mtest: Arguments have changed, and function has been improved.
    • plotxy: The number of methods have been reduced (for simplification).

    The aggregated syntax (matrice with two columns as input {x,y}) is not allowed anymore.

Version 0.2.2

  • New

    • @head: macro for function 'head'.

  • Modifications

    • Code improvement.

Version 0.2.1

  • New

    • splsrda: Sparse PLSR-DA.
    • splslda: Sparse PLS-LDA.
    • splsqda: Sparse PLS-QDA.
    • splskdeda: Sparse PLS-KDE-DA.

  • Modifications

    • Code improvement.

Version 0.2.0

  • New

    • splskern: Sparse PLSR.

  • Modifications

    • Code improvement.

Version 0.1.24

  • New

    • colmeanskip, colstdskip, colsumskip, colvarskip: Column-wise operations allowing missing data.
    • nsc: Nearest shrunken centroids (NSC).
    • nscda: Discrimination by nearest shrunken centroids.
    • pcanipals: PCA by NIPALS algorithm.
    • pcanipalsmiss: PCA by NIPALS algorithm allowing missing data.
    • plist Print each element of a list.
    • rowmeanskip, rowstdskip, rowsumskip, rowvarskip: Row-wise operations allowing missing data.
    • soft: Soft thresholding.
    • softmax: Softmax function.
    • spca, snipals, snipalsh, snipalsmix: Sparse PCA (Shen & Hunag 2008).

  • Modifications

    • Code improvement.

  • Breaking changes

    • covsel has been removed and integrated into covselr that was extended and made faster.

Version 0.1.23

  • New

    • dmnormlog: Logarithm of the normal probability density estimation.
    • rda: Regularized discriminant analysis (RDA)

  • Modifications

    • matB, matW, lda, qda: new argument 'weights'.
    • qda, plsqda, lwplsqda, plsqdaavg, lwplsqdaavg: new argument 'alpha' (continuum from QDA toward LDA).
    • dmnorm: new argument 'simpl'.
    • Code improvement.

  • Breaking changes

    • checkdupl, checkmiss: renamed to dupl and miss.
    • fda, fdasvd : argument 'pseudo' has been replaced by 'lb' (ridge regularization).

Version 0.1.22

  • New

    • difmean : Compute a detrimental matrix (for calibration transfer) by column means difference.

  • Modifications

    • Code improvement.

  • Breaking changes

    • calds, calpds: Order of arguments 'Xt' and 'X' were inverted (required to be useable by gridscore etc.).

Version 0.1.21

  • New

    • kdeda: Discriminant analysis using non-parametric kernel Gaussian density estimation (KDE-DA).
    • plskdeda: PLS-KDE-DA

  • Modifications

    • Code improvement.

  • Breaking changes

    • lda, qda: returned output 'ds' renamed to 'dens'.

Version 0.1.20

  • New

    • dmkern: Gaussian kernel density estimation (KDE)..
    • out: Return if elements of a vector are strictly outside of a given range.
    • pval: Compute p-value(s) for a distribution, an ECDF or a vector.

  • Modifications

    • plotxy: accept a matrix (n, 2) as input.
    • Code improvement.

  • Breaking changes

    • dens: removed and replaced by dmkern.

Version 0.1.19

  • New

  • Modifications

    • Dependance to unused package HypothesisTests.jl was removed.
    • Code cleaning.

Version 0.1.18

  • New

    • confusion Confusion matrix.
    • plotconf Plot confusion matrix.

  • Modifications

    • cplsravg: new argument 'typda'.
    • Code cleaning.

Version 0.1.17

  • New

    • lwmlrda: k-Nearest-Neighbours locally weighted MLR-based discrimination (kNN-LWMLR-DA).
    • lwmlrda_s: kNN-LWMLR-DA after preliminary (linear or non-linear) dimension reduction.
    • lwplsrda_s kNN-LWPLSR-DA after preliminary (linear or non-linear) dimension reduction.

  • Modifications

    • lwmlr_s: Add arguments 'psamp' and 'samp' for large nb. observations.
    • Code cleaning.

  • Breaking changes

    • lwplsr_s: Arguments and pipeline changed to be consistent with lwmlr_s.
    • sampclas: remamed to sampcla.

Version 0.1.16

  • New

    • dkplsrda Discrimination based on direct kernel partial least squares regression (DKPLSR-DA)
    • treer_dt Regression tree (CART) with DecisionTree.jl
    • rfr_dt Random forest regression with DecisionTree.jl
    • treeda_dt Discrimination tree (CART) with DecisionTree.jl
    • rfda_dt Random forest discrimination with DecisionTree.jl

  • Modifications

    • selwold : add argument 'step'-
    • Code cleaning.

  • Breaking changes

    • Warning: Difficult breaking bugs appeared in C++ dll from Julia v1.8.4 (still present in v1.9-betas) that removed the possibility to use packages LIBSVM.jl and XGBoost.jl under Windows. For this reason, Jchemo.jl stopped to use these two packages. All the related functions (SVM, RF and XGBoost models) were removed. For CART models (trees), they were replaced by new functions using package DecisionTree.jl.

Version 0.1.15

  • New

    • cosm Cosinus between the columns of a matrix
    • cosv Cosinus between two vectors
    • lwmlr: k-Nearest-neighbours locally weighted multiple linear regression (kNN-LWMLR)
    • lwmlr_s: kNN-LWMLR after preliminary (linear or non-linear) dimension reduction
    • pmod Short-cut for function 'Base.parentmodule'
    • tabdupl Tabulate duplicated values in a vector

  • Modifications

    • Improvement of vi_baggr
    • Code cleaning.

Version 0.1.14

  • New

    • isel: Interval variable selection.
    • mlev: Return the sorted levels of a dataset.
    • pcasph: Spherical PCA.
    • tabdf: Compute the nb. occurences of groups in categorical variables of a dataset.
    • vip: Variable importance by permutation.

  • Modifications

    • plotgrid: add of argument 'leg'.
    • plotxy: add of arguments 'circle' and 'zeros'.
    • Code cleaning.

  • Breaking changes

    • aggstat has changed (arguments).
    • baggr_vi renamed to vi_baggr
    • baggr_oob renamed to oob_baggr
    • gridcv and gridcv_mb: in output 'res_rep', colum 'rept' replaced by column 'repl'.
    • iplsr was removed and replaced by the more generic function isel.
    • mtest: Outputs 'idtrain' and 'idtest' renamed to 'train' and 'test'.
    • rd: argument 'corr' chanfed to 'typ'.
    • tabn was removed.
    • vimp_xgb renamed to vi_xgb
    • vip: outputs have been improved.

Version 0.1.13

  • New

    • mtest Select indexes defining training and test sets for each column of a dataframe.

  • Modifications

    • Code cleaning.

Version 0.1.12

  • Modifications
    • All tree functions: Internal changes to adapt to modifications in XGBoost.jl library.

Version 0.1.11

  • New
    • vip Variable importance on PLS projections (VIP).

Version 0.1.10

  • New

    • mbwcov: Multiblock weighted covariate analysis regression (MBWCov) (Mangana et al. 2021).

  • Modifications

    • Code cleaning.

  • Breaking changes

    • mbmang renamed to mbunif (Unified multiblock analysis).
    • ramang renamed to rrr (Reduced rank regression).

Version 0.1.9

  • New

    • rasvd: Redundancy analysis - PCA on instrumental variables (PCAIV).
    • ramang Redundancy analysis regression = Reduced rank regression (RRR)
    • mbplswest MBPLSR - Nipals algorithm (Westerhuis et al. 1998)

  • Breaking changes

    • All the functions ..._avg and ..._stack renamed to ...avg and ...stack (e.g. plsr_avg to plsravg).
    • caltransf_ds and caltransf_pds remaned to calds and calpds.
    • fnorm renamed to frob.

Version 0.1.8

  • New

    • plswold: PLSR Wold Nipals algorithm.
    • ccawold: CCA Wold Nipals algorithm.
    • mbmang: Unified multiblock data analysis of Mangana et al. 2019.

  • Modifications

    • mlrpinv_n renamed to mlrpinvn
    • pls renamed to plscan.
    • pls_svd renamed to plstuck.
    • rcca renamed to cca (and argument 'alpha" to 'tau').
    • rpmat_gauss and rpmat_li renamed to rpmatgauss and rpmatli
    • Output 'Tbl' added in comdim and mbpca.
    • Code cleaning.

Version 0.1.7

  • CairoMakie.jl was removed from the dependances, and replaced by Makie.jl.

To display the plots, the user has to install and load one of the Makie's backend (e.g. CairoMakie).

  • New

    • rcca: Canonical correlation analysis. (RCCA).
    • pls: Canonical partial least squares regression (Symmetric PLS).
    • pls_svd: Tucker's inter-battery method of factor analysis (PLS-SVD).
    • colnorm2 was removed, replaced by colnorm:

    Norm of each column of a dataset.

    • fnorm: Frobenius norm of a matrix.
    • norm2 was removed, replaced by normw:

    Weighted norm of a vector.

  • Modifications

    • Major changes in multiblock functions:
      • Renamed functions:
        • mbpca_cons ==> mbpca
        • mbpcacomdims ==> comdim
        • mbplsr_rosa ==> rosaplsr
        • mbplsr_so ==> soplsr
      • Argument 'X_bl' renamed to 'Xbl'
    • Variable 'pc' in summary outputs of PCA and KPCA functions renamed to 'lv'.
    • Modification of all the tree functions to adapt to the new version of

    XGBoost.jl (>= 2.02) (https://juliahub.com/ui/Packages/XGBoost/rSeEh/2.0.2). The new Jchemo functions does not work anymore with XGBoost.jl 1.5.2.

    • Code cleaning.

Version 0.1.6

  • Package Jchemo.jl has been registered.

  • Modifications

    • Code cleaning.

Version 0.1.5

  • Modifications
    • Code cleaning.

Version 0.1.4

  • New

    • head: Display the first rows of a dataset.

  • Modifications

    • Remove of side-effects in some functions of multi-bloc analyses.

Version 0.1.3

  • Modifications
    • detrend: argument 'degree' renamed to 'pol'.
    • Code cleaning.

Version 0.1.2

  • Modifications
    • detrend: new argument 'degree'
    • gridcvlv: correction of a bug (typing error) inserted in the last version.

Version 0.1.1

  • Modifications
    • blockscal: bug corrected in arguments.
    • Use of multi-threading (package Threads) in functions locw and locwlv, used in local models.

Version 0.1.0

  • Modifications
    • Argument 'scal' (X and/or Y column-scaling) added to various functions.
    • blockscal: names of arguments have changed.
    • plotgrid: argument 'indx' modified.

Version 0.0.26

  • New
    • cscale: Center and scale each column of a matrix.
  • Modifications
    • Argument 'scal' (X and/or Y column-scaling) added to various functions. Work in progress. The argument will be available for all the concerned fonctions.
    • Output 'explvar' replaced by 'explvarx' in all the concerned functions.
    • rd: New argument 'corr'.

Version 0.0.25

  • New

    • rd: Redundancy coefficients between two matrices.

  • Modifications

    • summary for Plsr objects. See the example in ?plskern.

Version 0.0.24

  • Modifications
    • selwold: Argument "plot" renamed "graph" and bug fixed in plotting.

Version 0.0.23

  • New

    • occknndis: One-class classification using "global" k-nearest neighbors distances.
    • occlknndis: One-class classification using "local" k-nearest neighbors distances.

  • Modifications

    • occsd, occod, occsdod, occstah: The methods to compute the cutoff have changed.

Version 0.0.22

  • New

    • colmad: Median absolute deviation (MAD) of each column of a matrix.
    • occsdod: One-class classification using a compromise between PCA/PLS score (SD) and orthogonal (OD) distances.
    • replacedict: Replace the elements of a vector by levels defined in a dictionary.
    • stah: Stahel-Donoho outlierness measure.

  • Modifications

    • dens: outputs have been modified.
    • odis and scordis have been rename to occsd and occod, and modified.
    • plotxy: new argument "bisect".

Version 0.0.21

  • New

    • dens: Univariate kernel density estimation.

  • Modifications

    • All the datasets (examples) have been moved to package JchemoData (https://github.com/mlesnoff/JchemoData.jl)
    • plotsp: Argument 'nsamp' added.
    • datasets: removed and transferred to JchemoData.jl

Version 0.0.20

  • New

    • covselr: Covsel regression (Covsel+Mlr).

  • Modifications

    • covsel, mlrvec: Arguments changed.

Version 0.0.19

  • New

    • selwold : Wold's criterion to select dimensionality in LV (e.g. PLSR) models.
    • plotxy : Scatter plot (x, y) data.

  • Modifications

    • plotscore: Renamed to plotgrid.

Version 0.0.18

  • New
    • plotgrid : Plot error or performance rates of model predictions.
  • Modifications
    • plotsp: argument 'size' was added.

Version 0.0.17

  • New

    • replacebylev2 : Replace the elements of an index-vector by levels.

  • Modifications

    • aggstat : Sorting order for dataframes.
    • checkdupl : bug corrected.
    • matB, matW : when requested, update of covm to cov, and aggstat output.
    • plotsp : faster.
    • transfer_ds : renamed to caltransf_ds.
    • transfer_pds : renamed to caltransf_pds.
    • recodcat2num : renamed to recodcat2int
    • segmts : A seed (MersenneTwister) can be set for the random samplings.
    • Examples added in the helps of every functions.
    • Discrimination functions: major updates.

Version 0.0.16

  • New

    • transfer_ds : Calibration transfert with direct standardization (DS).
    • transfer_pds : Calibration transfert with piecewise direct standardization (PDS).

  • Modifications

    • mlr functions : Argument 'noint' added.
    • plsravgcv : Bug corrected.

Version 0.0.15

  • New

    • plsr_stack : Stacking PLSR models

  • Modifications

    • aicplsr : BIC criterion added
    • fweight
    • plsr_avg : Stacking was added
    • plsravgaic
    • plsravgcv
    • lwplsr_avg

Version 0.0.14

  • New
    • lwplsr_s

Version 0.0.13

  • New

    • fweight
    • rowmean, rowstd

  • Modifications

    • aicplsr
    • lwplsr_avg
    • plsr_avg
    • snv
    • wshenk

Version 0.0.12

  • New

    • nco, nro

  • Modifications

    • mpars renamed to mpar
    • All functions terminating with "...agg" renamed to "...avg".

Version 0.0.11

  • New

    • blockscal_mfa
    • datasets
    • mbpca_cons
    • lg
    • ssq

  • Modifications

    • All the functions terminating with a "s" have been renamed without "s"

    (e.g. colmeans was renamed to colmean)

Version 0.0.10

  • New

    • colsum
    • mbpcacomdims
    • rowsum

  • Modifications

    • nipals
    • mse
    • mbpls

Version 0.0.9

  • New

    • blockscal_frob, blockscal_ncol, blockscal_sd
    • colnorm2
    • corm, covm
    • nipals
    • norm2

  • Modifications

    • blockscal
    • matcov renamed to covm and extended

  • Removed

    • mbplsrmidavg
    • mbplsr_mid
    • mbplsrmidseq

Version 0.0.8

  • New functions
    • gridcv_mb
    • gridcvlv_mb
    • mbplsr_avg
    • mbplsr_mid
    • mbplsrmidseq
  • Modifications
    • rosaplsr renamed to mbplsr_rosa
    • soplsr renamed to mbplsr_soplsr

Version 0.0.7

  • New functions
    • mbplsr
    • soplsr

Version 0.0.6

  • New functions
    • colstd
    • plsrosa
    • plssimp
    • rosaplsr
    • rv
    • rmrows, rmcols: renamed to rmrow, rmcol
  • Modifications
    • interpl, interpl_mon: changes in arguments
    • plotsp: changes in outputs
    • aggstat (::AbstractMatrix): changes in arguments and outputs

Version 0.0.5

  • New functions
    • blockscal
    • pcr
    • rp
    • rpmatgauss
    • rpmat_li

Version 0.0.4

  • New functions

    • iplsr

  • Modification of covsel

Version 0.0.3

  • New functions
    • interpl
    • checkdupl, checkmiss

Version 0.0.2

  • New functions
    • covsel
    • interpl has been replaced by interpl_mon
  • Change in output of vi_xgb

Version 0.0.1

First version of the package