Jchemo was initially dedicated to partial least squares regression (PLSR) and discrimination (PLSDA) and their extensions, in particular locally weighted PLS models (KNN-LWPLS-R & -DA; e.g. https://doi.org/10.1002/cem.3209). In a second phase, the package has expand to many other models for dimension reduction and regression and discrimination (see the list of functions here).
Why the name Jchemo?: Since it is orientated to chemometrics , in brief the use of biometrics for chemistry. But most of the provided methods are generic to other types of data than chemistry.
Jchemo is organized between
- transform operators (that have a function
transf), - predictors (that have a function
predict), - utility functions.
Some models, such as PLSR models, are both a transform operator and a predictor.
Ad'hoc pipelines of operations can also be built. In Jchemo, a pipeline is a chain of K modeling steps containing
- either K transform steps,
- or K - 1 transform steps and a final prediction step.
The pipelines are built with function pip, see here.
To install Jchemo
- From the official Julia repo, run in the Pkg REPL:
pkg> add Jchemoor for a specific version, for instance:
pkg> add Jchemo@0.1.18- For the current developing version (potentially not stable):
pkg> add https://github.com/mlesnoff/Jchemo.jl.gitBefore to update the package, it is recommended to have a look on What changed to avoid eventual problems due to breaking changes.
Most the Jchemo functions have keyword arguments (kwargs). The keyword arguments required/allowed in a function can be found in the Index of function section of the documentation:
or in the REPL by displaying the function's help page, for instance for function plskern:
julia> ?plskernThe kwargs arguments and their default values are defined in containers depending of the functions, available
here. For a given function, they can be displayed in the REPL with function default, e.g.:
julia> default(plskern)
Jchemo.ParPlsr
nlv: Int64 1
scal: Bool falseExamples of syntax are given at the end of this README, and in the help pages of the functions.
The datasets used in the help page examples are stored in the package JchemoData.jl, a repository of datasets on chemometrics and other domains. Examples of scripts demonstrating Jchemo are also available in the pedagogical project JchemoDemo.
Two generic grid-search functions are available to tune the predictors:
Highly accelerated versions of these tuning tools have been implemented for models based on latent variables (LVs) and ridge regularization.
Some Jchemo functions (e.g. those using kNN selections) use multi-threading to speed the computations. Taking advantage of this requires to specify a relevant number of threads (for instance from the Settings menu of the VsCode Julia extension and the file settings.json).
Jchemo uses Makie for plotting. Displaying the plots requires to install and load one of the Makie's backends (e.g. CairoMakie).
using Jchemo, BenchmarkToolsjulia> versioninfo()
Julia Version 1.10.0
Commit 3120989f39 (2023-12-25 18:01 UTC)
Build Info:
Official https://julialang.org/ release
Platform Info:
OS: Windows (x86_64-w64-mingw32)
CPU: 16 × Intel(R) Core(TM) i9-10885H CPU @ 2.40GHz
WORD_SIZE: 64
LIBM: libopenlibm
LLVM: libLLVM-15.0.7 (ORCJIT, skylake)
Threads: 23 on 16 virtual cores
Environment:
JULIA_EDITOR = coden = 10^6 # nb. observations (samples)
p = 500 # nb. X-variables (features)
q = 10 # nb. Y-variables to predict
nlv = 25 # nb. PLS latent variables
X = rand(n, p)
Y = rand(n, q)
zX = Float32.(X)
zY = Float32.(Y)## Float64
## (NB.: multi-threading is not used in plskern)
model = plskern(; nlv)
@benchmark fit!($model, $X, $Y)
BenchmarkTools.Trial: 1 sample with 1 evaluation.
Single result which took 7.532 s (1.07% GC) to evaluate,
with a memory estimate of 4.09 GiB, over 2677 allocations.# Float32
@benchmark fit!($model, $zX, $zY)
BenchmarkTools.Trial: 2 samples with 1 evaluation.
Range (min … max): 3.956 s … 4.148 s ┊ GC (min … max): 0.82% … 3.95%
Time (median): 4.052 s ┊ GC (median): 2.42%
Time (mean ± σ): 4.052 s ± 135.259 ms ┊ GC (mean ± σ): 2.42% ± 2.21%
█ █
█▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁█ ▁
3.96 s Histogram: frequency by time 4.15 s <
Memory estimate: 2.05 GiB, allocs estimate: 2677.n = 150 ; p = 200
q = 2 ; m = 50
Xtrain = rand(n, p)
Ytrain = rand(n, q)
Xtest = rand(m, p)
Ytest = rand(m, q) Let us consider a signal preprocessing with the Savitsky-Golay filter, using function savgol. The keyword arguments of savgol are npoint, deriv and degree. See for instance in the REPL:
julia> ?savgolThe syntax to fit the model is as follows:
## Below, the order of the kwargs is not
## important but the argument names have
## to be correct.
## Keywords arguments are specified
## after character ";"
## Model definition
## (below, the name 'model' can be replaced by any other name)
npoint = 11 ; deriv = 2 ; degree = 3
model = savgol(; npoint, deriv, degree)
## Fitting
fit!(model, Xtrain)which is the strictly equivalent to:
## Below, ";" is not required since the kwargs values are
## specified within the function
model = savgol(npoint = 11, deriv = 2, degree = 3)
fit!(model, Xtrain)Contents of object model can be displayed by:
julia> pnames(model)
(:algo, :fitm, :kwargs)Sub-objects algo, fitm and kwargs contain the used algorithm (a function), the model fitted on the data, and the kwargs arguments, respectively.
Once the model is fitted, the transformed data are given by:
Xptrain = transf(model, Xtrain) # preprocessed data
Xptest = transf(model, Xtest)Several preprocessing can be applied sequentially to the data by building a pipeline (see section Fitting a pipeline thereafter for examples).
Let us consider a principal component analysis (PCA), using function pcasvd.
The syntax to fit the model using a SVD decomposition is as follows:
nlv = 15 # nb. principal components
model = pcasvd(; nlv)
fit!(model, Xtrain, ytrain)The PCA score matrices (i.e. the projections of the observations on the PCA directions) can be computed by:
Ttrain = transf(model, Xtrain)
Ttest = transf(model, Xtest)Object Ttrain above can also be obtained directly by:
Ttrain = model.fitm.TSome model summary (% of explained variance, etc.) can be displayed by:
summary(model, Xtrain)For a preliminary scaling of the data before the PCA decomposition, the syntax is:
nlv = 15 ; scal = true
model = pcasvd(; nlv, scal)
fit!(model, Xtrain, ytrain)Let us consider a Gaussian kernel partial least squares regression (KPLSR), using function kplsr.
The syntax to fit the model is as follows:
nlv = 15 # nb. latent variables
kern = :krbf ; gamma = .001
model = kplsr(; nlv, kern, gamma)
fit!(model, Xtrain, ytrain)As for PCA, the score matrices can be computed by:
Ttrain = transf(model, Xtrain) # = model.fitm.T
Ttest = transf(model, Xtest)and model summary by:
summary(model, Xtrain)Y-Predictions are given by:
pred = predict(model, Xtest).predExamples of tuning of predictive models (test-set validation and K-fold cross-validation) are given in the help pages of functions gridscore and gridcv:
?gridscore
?gridcvLet us consider a data preprocessing by standard-normal-variation transformation (SNV) followed by a Savitsky-Golay filter and a polynomial de-trending transformation.
The pipeline is fitted as follows:
## Models' definition
model1 = snv()
model2 = savgol(npoint = 5, deriv = 1, degree = 2)
model3 = detrend_pol()
## Pipeline building
model = pip(model1, model2, model3)
## Fitting
fit!(model, Xtrain)The transformed data are given by:
Xptrain = transf(model, Xtrain)
Xptest = transf(model, Xtest)Let us consider a support vector machine regression model implemented on preliminary computed PCA scores (PCA-SVMR).
The pipeline is fitted as follows:
nlv = 15
kern = :krbf ; gamma = .001 ; cost = 1000
model1 = pcasvd(; nlv)
model2 = svmr(; kern, gamma, cost)
model = pip(model1, model2)
fit!(model, Xtrain)The Y-predictions are given by:
pred = predict(model, Xtest).predAny step(s) of data preprocessing can obviously be implemented before the modeling, either outside of the given predictive pipeline or being involded directlty in the pipeline, such as for instance:
degree = 2 # de-trending with polynom degree 2
nlv = 15
kern = :krbf ; gamma = .001 ; cost = 1000
model1 = detrend_pol(; degree)
model2 = pcasvd(; nlv)
model3 = svmr(; kern, gamma, cost)
model = pip(model1, model2, model3)The LWR algorithm of Naes et al (1990) consists in implementing a preliminary global PCA on the data and then a kNN locally weighted multiple linear regression (kNN-LWMLR) on the global PCA scores.
The pipeline is defined by:
nlv = 25
metric = :eucl ; h = 2 ; k = 200
model1 = pcasvd(; nlv)
model2 = lwmlr(; metric, h, k)
model = pip(model1, model2)Naes et al., 1990. Analytical Chemistry 664–673.
The pipeline of Shen et al. (2019) consists in implementing a preliminary global PLSR on the data and then a kNN-PLSR on the global PLSR scores.
The pipeline is defined by:
nlv = 25
metric = :mah ; h = Inf ; k = 200
model1 = plskern(; nlv)
model2 = lwplsr(; metric, h, k)
model = pip(model1, model2)Shen et al., 2019. Journal of Chemometrics, 33(5) e3117.
Matthieu Lesnoff
contact: [email protected]
-
Cirad, UMR Selmet, Montpellier, France
-
ChemHouse, Montpellier
Lesnoff, M. 2021. Jchemo: Chemometrics and machine learning on high-dimensional data with Julia. https://github.com/mlesnoff/Jchemo. UMR SELMET, Univ Montpellier, CIRAD, INRA, Institut Agro, Montpellier, France
- G. Cornu (Cirad) https://ur-forets-societes.cirad.fr/en/l-unite/l-equipe
- M. Metz (Pellenc ST, Pertuis, France)
- L. Plagne, F. Févotte (Triscale.innov) https://www.triscale-innov.com
- R. Vezy (Cirad) https://www.youtube.com/channel/UCxArXLI-gxlTmWGGgec5D7w