From version 0.7.3, I started getting an error ERROR: cannot specify both lambda and lambda_min_ratio when fitting wide model matrices to binomial response data and specifying lambda (but not lambda_min_ratio).
I think the reason is this line of code, where a length(y) was changed to a size(y, 1) in version 0.7.3 - which is not the same thing when fitting binomial data.
MWE:
n = 1000
m = 1200
y = rand(Bool, n)
y_mat = [(1 .- y) y]
X = rand(n, m)
lambda = collect(0.01:0.01:1)
glmnet(X, y_mat, GLMNet.Binomial(); lambda)
From version 0.7.3, I started getting an error
ERROR: cannot specify both lambda and lambda_min_ratiowhen fitting wide model matrices to binomial response data and specifying lambda (but not lambda_min_ratio).I think the reason is this line of code, where a
length(y)was changed to asize(y, 1)in version 0.7.3 - which is not the same thing when fitting binomial data.MWE: