Classification using the artificial neural network algorithm.
Usage
nnetClassification(
object,
assessRes,
scores = c("prediction", "all", "none"),
decay,
size,
fcol = "markers",
...
)Arguments
- object
An instance of class
"MSnSet".- assessRes
An instance of class
"GenRegRes", as generated bynnetOptimisation.- scores
One of
"prediction","all"or"none"to report the score for the predicted class only, for all classes or none.- decay
If
assessResis missing, adecaymust be provided.- size
If
assessResis missing, asizemust be provided.- fcol
The feature meta-data containing marker definitions. Default is
markers.- ...
Additional parameters passed to
nnetfrom packagennet.
Value
An instance of class "MSnSet" with
nnet and nnet.scores feature variables storing
the classification results and scores respectively.
Examples
library(pRolocdata)
data(dunkley2006)
## reducing parameter search space and iterations
params <- nnetOptimisation(dunkley2006, decay = 10^(c(-1, -5)), size = c(5, 10), times = 3)
#>
|
| | 0%
|
|===== | 7%
|
|========= | 13%
|
|============== | 20%
|
|=================== | 27%
|
|======================= | 33%
|
|============================ | 40%
|
|================================= | 47%
|
|===================================== | 53%
|
|========================================== | 60%
|
|=============================================== | 67%
|
|=================================================== | 73%
|
|======================================================== | 80%
|
|============================================================= | 87%
|
|================================================================= | 93%
|
|======================================================================| 100%
params
#> Object of class "GenRegRes"
#> Algorithm: nnet
#> Hyper-parameters:
#> decay: 0.1 1e-05
#> size: 5 10
#> Design:
#> Replication: 3 x 5-fold X-validation
#> Partitioning: 0.2/0.8 (test/train)
#> Results
#> macro F1:
#> Min. 1st Qu. Median Mean 3rd Qu. Max.
#> 0.9645 0.9681 0.9718 0.9736 0.9782 0.9846
#> best decay: 1e-05 0.1
#> best size: 10
plot(params)
f1Count(params)
#> 10
#> 1e-05 0
#> 0.1 1
levelPlot(params)
getParams(params)
#> decay size
#> 0.1 10.0
res <- nnetClassification(dunkley2006, params)
#> [1] "markers"
#> # weights: 269
#> initial value 650.117805
#> iter 10 value 376.359754
#> iter 20 value 238.559795
#> iter 30 value 178.672209
#> iter 40 value 171.805580
#> iter 50 value 162.533681
#> iter 60 value 160.271625
#> iter 70 value 158.637365
#> iter 80 value 157.563421
#> iter 90 value 156.799485
#> iter 100 value 156.548913
#> final value 156.548913
#> stopped after 100 iterations
getPredictions(res, fcol = "nnet")
#> ans
#> ER lumen ER membrane Golgi Mitochondrion PM
#> 18 186 94 106 131
#> Plastid Ribosome TGN vacuole
#> 49 53 20 32
#> MSnSet (storageMode: lockedEnvironment)
#> assayData: 689 features, 16 samples
#> element names: exprs
#> protocolData: none
#> phenoData
#> sampleNames: M1F1A M1F4A ... M2F11B (16 total)
#> varLabels: membrane.prep fraction replicate
#> varMetadata: labelDescription
#> featureData
#> featureNames: AT1G09210 AT1G21750 ... AT4G39080 (689 total)
#> fvarLabels: assigned evidence ... nnet.pred (11 total)
#> fvarMetadata: labelDescription
#> experimentData: use 'experimentData(object)'
#> pubMedIds: 16618929
#> Annotation:
#> - - - Processing information - - -
#> Loaded on Thu Jul 16 22:53:08 2015.
#> Normalised to sum of intensities.
#> Added markers from 'mrk' marker vector. Thu Jul 16 22:53:08 2015
#> Performed nnet prediction (decay=0.1 size=10) Sat Nov 23 16:04:25 2024
#> Added nnet predictions according to global threshold = 0 Sat Nov 23 16:04:25 2024
#> MSnbase version: 1.17.12
getPredictions(res, fcol = "nnet", t = 0.75)
#> ans
#> ER lumen ER membrane Golgi Mitochondrion PM
#> 14 148 68 93 95
#> Plastid Ribosome TGN unknown vacuole
#> 42 35 13 154 27
#> MSnSet (storageMode: lockedEnvironment)
#> assayData: 689 features, 16 samples
#> element names: exprs
#> protocolData: none
#> phenoData
#> sampleNames: M1F1A M1F4A ... M2F11B (16 total)
#> varLabels: membrane.prep fraction replicate
#> varMetadata: labelDescription
#> featureData
#> featureNames: AT1G09210 AT1G21750 ... AT4G39080 (689 total)
#> fvarLabels: assigned evidence ... nnet.pred (11 total)
#> fvarMetadata: labelDescription
#> experimentData: use 'experimentData(object)'
#> pubMedIds: 16618929
#> Annotation:
#> - - - Processing information - - -
#> Loaded on Thu Jul 16 22:53:08 2015.
#> Normalised to sum of intensities.
#> Added markers from 'mrk' marker vector. Thu Jul 16 22:53:08 2015
#> Performed nnet prediction (decay=0.1 size=10) Sat Nov 23 16:04:25 2024
#> Added nnet predictions according to global threshold = 0.75 Sat Nov 23 16:04:25 2024
#> MSnbase version: 1.17.12
plot2D(res, fcol = "nnet")
