As a note beforehand, I think I should mention that I am working with highly sensitive medical data that is protected by HIPAA. I cannot share real data with dput- it would be illegal to do so. That is why I made a fake dataset and explained my processes to help reproduce the error.
I have been trying to estimate an elastic net model in r using glmnet. However, I keep getting an error. I am not sure what is causing it. The error happens when I go to train the data. It sounds like it has something to do with the data type and matrix.
I have provided a sample dataset. Then I set the outcomes and certain predictors to be factors. After setting certain variables to be factors, I label them. Next, I create an object with the column names of the predictors I want to use. That object is pred.names.min. Then I partition the data into the training and test data frames. 65% in the training, 35% in the test. With the train control function, I specify a few things I want to have happen with the model- random paraments for lambda and alpha, as well as the leave one out method. I also specify that it is a classification model (categorical outcome). In the last step, I specify the training model. I write my code to tell it to use all of the predictor variables in the pred.names.min object for the trainingset data frame.
library(dplyr)
library(tidyverse)
library(glmnet),0,1,0
library(caret)
#creating sample dataset
df<-data.frame("BMIfactor"=c(1,2,3,2,3,1,2,1,3,2,1,3,1,1,3,2,3,2,1,2,1,3),
"age"=c(0,4,8,1,2,7,4,9,9,2,2,1,8,6,1,2,9,2,2,9,2,1),
"L_TartaricacidArea"=c(0,1,1,0,1,1,1,0,0,1,0,1,1,0,1,0,0,1,1,0,1,1),
"Hydroxymethyl_5_furancarboxylicacidArea_2"=
c(1,1,0,1,0,0,1,0,1,1,0,1,1,0,1,1,0,1,0,1,0,1),
"Anhydro_1.5_D_glucitolArea"=
c(8,5,8,6,2,9,2,8,9,4,2,0,4,8,1,2,7,4,9,9,2,2),
"LevoglucosanArea"=
c(6,2,9,2,8,6,1,8,2,1,2,8,5,8,6,2,9,2,8,9,4,2),
"HexadecanolArea_1"=
c(4,9,2,1,2,9,2,1,6,1,2,6,2,9,2,8,6,1,8,2,1,2),
"EthanolamineArea"=
c(6,4,9,2,1,2,4,6,1,8,2,4,9,2,1,2,9,2,1,6,1,2),
"OxoglutaricacidArea_2"=
c(4,7,8,2,5,2,7,6,9,2,4,6,4,9,2,1,2,4,6,1,8,2),
"AminopentanedioicacidArea_3"=
c(2,5,5,5,2,9,7,5,9,4,4,4,7,8,2,5,2,7,6,9,2,4),
"XylitolArea"=
c(6,8,3,5,1,9,9,6,6,3,7,2,5,5,5,2,9,7,5,9,4,4),
"DL_XyloseArea"=
c(6,9,5,7,2,7,0,1,6,6,3,6,8,3,5,1,9,9,6,6,3,7),
"ErythritolArea"=
c(6,7,4,7,9,2,5,5,8,9,1,6,9,5,7,2,7,0,1,6,6,3),
"hpresponse1"=
c(1,0,1,1,0,1,1,0,0,1,0,0,1,0,1,1,1,0,1,0,0,1),
"hpresponse2"=
c(1,0,1,0,0,1,1,1,0,1,0,1,0,1,1,0,1,0,1,0,0,1))
#setting variables as factors
df$hpresponse1<-as.factor(df$hpresponse1)
df$hpresponse2<-as.factor(df$hpresponse2)
df$BMIfactor<-as.factor(df$BMIfactor)
df$L_TartaricacidArea<- as.factor(df$L_TartaricacidArea)
df$Hydroxymethyl_5_furancarboxylicacidArea_2<-
as.factor(df$Hydroxymethyl_5_furancarboxylicacidArea_2)
#labeling factor levels
df$hpresponse1 <- factor(df$hpresponse1, labels = c("group1.2", "group3.4"))
df$hpresponse2 <- factor(df$hpresponse2, labels = c("group1.2.3", "group4"))
df$L_TartaricacidArea <- factor(df$L_TartaricacidArea, labels =c ("No",
"Yes"))
df$Hydroxymethyl_5_furancarboxylicacidArea_2 <-
factor(df$Hydroxymethyl_5_furancarboxylicacidArea_2, labels =c ("No",
"Yes"))
df$BMIfactor <- factor(df$BMIfactor, labels = c("<40", ">=40and<50",
">=50"))
#creating list of predictor names
pred.start.min <- which(colnames(df) == "BMIfactor"); pred.start.min
pred.stop.min <- which(colnames(df) == "ErythritolArea"); pred.stop.min
pred.names.min <- colnames(df)[pred.start.min:pred.stop.min]
#partition data into training and test (65%/35%)
set.seed(2)
n=floor(nrow(df)*0.65)
train_ind=sample(seq_len(nrow(df)), size = n)
trainingset=df[train_ind,]
testingset=df[-train_ind,]
#specifying that I want to use the leave one out cross-
#validation method and
use "random" as search for elasticnet
tcontrol <- trainControl(method = "LOOCV",
search="random",
classProbs = TRUE)
#training model
elastic_model1 <- train(as.matrix(trainingset[,
pred.names.min]),
trainingset$hpresponse1,
data = trainingset,
method = "glmnet",
trControl = tcontrol)
After I run the last chunk of code, I end up with this error:
Error in { :
task 1 failed - "error in evaluating the argument 'x' in selecting a
method for function 'as.matrix': object of invalid type "character" in
'matrix_as_dense()'"
In addition: There were 50 or more warnings (use warnings() to see the first
50)
I tried removing the "as.matrix" arguemtent:
elastic_model1 <- train((trainingset[, pred.names.min]),
trainingset$hpresponse1,
data = trainingset,
method = "glmnet",
trControl = tcontrol)
It still produces a similar error.
Error in { :
task 1 failed - "error in evaluating the argument 'x' in selecting a method
for function 'as.matrix': object of invalid type "character" in
'matrix_as_dense()'"
In addition: There were 50 or more warnings (use warnings() to see the first
50)
When I tried to make none of the predictors factors (but keep outcome as factor), this is the error I get:
Error: At least one of the class levels is not a valid R variable name; This
will cause errors when class probabilities are generated because the
variables names will be converted to X0, X1 . Please use factor levels that
can be used as valid R variable names (see ?make.names for help).
How can I fix this? How can I use my predictors (both the numeric and categorical ones) without producing an error?
glmnet does not handle factors well. The recommendation currently is to dummy code and re-code to numeric where possible:
Using LASSO in R with categorical variables
I am new to neural networks and the mxnet package in R. I want to do a logistic regression on my predictors since my observations are probabilities varying between 0 and 1. I'd like to weight my observations by a vector obsWeights I have, but I'm not sure where to implement the weights. There seems to be a weight= option in mx.symbol.FullyConnected but if I try weight=obsWeights I get the following error message
Error in mx.varg.symbol.FullyConnected(list(...)) :
Cannot find argument 'weight', Possible Arguments:
----------------
num_hidden : int, required
Number of hidden nodes of the output.
no_bias : boolean, optional, default=False
Whether to disable bias parameter.
How should I proceed to weight my observations? Here is my code at the moment.
# Prepare data
train.mm = model.matrix(obs ~ . , data = train_data)
train_label = train_data$obs
# Normalize
train.mm = apply(train.mm, 2, function(x) (x-min(x))/(max(x)-min(x)))
# Create MXDataIter compatible iterator
batch_size = 128
train.iter = mx.io.arrayiter(data=t(train.mm), label=train_label,
batch.size=batch_size, shuffle=T)
# Symbolic model definition
data = mx.symbol.Variable('data')
fc1 = mx.symbol.FullyConnected(data=data, num.hidden=128, name='fc1')
act1 = mx.symbol.Activation(data=fc1, act.type='relu', name='act1')
final = mx.symbol.FullyConnected(data=act1, num.hidden=1, name='final')
logistic = mx.symbol.LogisticRegressionOutput(data=final, name='logistic')
# Run model
mxnet_train = mx.model.FeedForward.create(
symbol = logistic,
X = train.iter,
initializer = mx.init.Xavier(rnd_type = 'gaussian', factor_type = 'avg', magnitude = 2),
num.round = 25)
Assigning the fully connected weight argument is not what you want to do at any rate. That weight is a reference to parameters of the layer; i.e., what you multiply in the inputs by to get output values These are the parameter values you're trying to learn.
If you want to make some samples matter more than others, then you'll need to adjust the loss function. For example, multiply the usual loss function by your weights so that they do not contribute as much to the overall average loss.
I do not believe the standard Mxnet loss functions have a spot for assigning weights (that is LogisticRegressionOutput won't cover this). However, you can make your own cost function that does. This would involve passing your final layer through a sigmoid activation function to first generate the usual logistic regression output value. Then pass that into the loss function you define. You could do squared error, but for logistic regression you'll probably want to use the cross entropy function:
l * log(y) + (1 - l) * log(1 - y),
where l is the label and y is the predicted value.
Ideally, you'd write a symbol with an efficient definition of the gradient (Mxnet has a cross entropy function, but its for softmax input, not a binary output. You could translate your output to two outputs with softmax as an alternative, but that seems less easy to work with in this case), but the easiest path would be to let Mxnet do its autodiff on it. Then you multiply that cross entropy loss by the weights.
I haven't tested this code, but you'd ultimately have something like this (this is what you'd do in python, should be similar in R):
label = mx.sym.Variable('label')
out = mx.sym.Activation(data=final, act_type='sigmoid')
ce = label * mx.sym.log(out) + (1 - label) * mx.sym.log(1 - out)
weights = mx.sym.Variable('weights')
loss = mx.sym.MakeLoss(weigths * ce, normalization='batch')
Then you want to input your weight vector into the weights Variable along with your normal input data and labels.
As an added tip, the output of an mxnet network with a custom loss via MakeLoss outputs the loss, not the prediction. You'll probably want both in practice, in which case its useful to group the loss with a gradient-blocked version of the prediction so that you can get both. You'd do that like this:
pred_loss = mx.sym.Group([mx.sym.BlockGrad(out), loss])