Error while trying to replicate a logistic regression problem using FluxML

Specific Domains Machine Learning
1

Hi,

I have been trying to replicate the code from the following link to understand how to use ML in Julia better
https://akaysh.github.io/Logistic-Regression-with-Julia/

While replicating I realized that when I ran the following code:

data = repeated((X, y), 200)
Flux.train!(loss, data, opt, cb = evalcb)

the result was this:
loss(X, y) = 0.6931471805599452 (tracked)
loss(X, y) = 0.6931471805599452 (tracked)
loss(X, y) = 0.6931471805599452 (tracked)
loss(X, y) = 0.6931471805599452 (tracked)
loss(X, y) = 0.6931471805599452 (tracked)
loss(X, y) = 0.6931471805599452 (tracked)
loss(X, y) = 0.6931471805599452 (tracked)
loss(X, y) = 0.6931471805599452 (tracked)
loss(X, y) = 0.6931471805599452 (tracked)
loss(X, y) = 0.6931471805599452 (tracked)
loss(X, y) = 0.6931471805599452 (tracked)
loss(X, y) = 0.6931471805599452 (tracked)
loss(X, y) = 0.6931471805599452 (tracked)
loss(X, y) = 0.6931471805599452 (tracked)
loss(X, y) = 0.6931471805599452 (tracked)
loss(X, y) = 0.6931471805599452 (tracked)
loss(X, y) = 0.6931471805599452 (tracked)
loss(X, y) = 0.6931471805599452 (tracked)
loss(X, y) = 0.6931471805599452 (tracked)
loss(X, y) = 0.6931471805599452 (tracked)
loss(X, y) = 0.6931471805599452 (tracked)
loss(X, y) = 0.6931471805599452 (tracked)
loss(X, y) = 0.6931471805599452 (tracked)
loss(X, y) = 0.6931471805599452 (tracked)
loss(X, y) = 0.6931471805599452 (tracked)

As you can see there is no change in the loss function.
Could anyone help me with what the problem is?

Thanks

2

@akaysh

3

Hi, im trying to do the same tutorial and I think the problem comes from the predict(x) function, when I’m writing it down in juno, I get the next error

StringIndexError(“predict(x) = NNlib.σ.(”, 21) string_index_err(::String, ::Int64) at string.jl:12
getindex(::String, ::UnitRange{Int64}) at string.jl:246
fuzzycompletionadapter(::String, ::String, ::String, ::Int64, ::Int64, ::Bool) at completions.jl:130
(::getfield(Atom, Symbol(“##279#280”)))(::Dict{String,Any}) at completions.jl:54
handlemsg(::Dict{String,Any}, ::Dict{String,Any}) at comm.jl:169
(::getfield(Atom, Symbol(“##31#33”)){Array{Any,1}})() at task.jl:259

Once I write the second point in the function

predict(x) = NNlib.σ.(x*W .+ b)
.-…-…-…-…-…-…^ Here

But don’t know how to solve it.

Hope bumping this thread someone could help us :pray:.

4

Looks like a unicode bug

Did you try the following

predict(x) = σ.(x*W .+ b)

?
That should work if you have Flux in scope.

What OS are you using?

5

Thanks, but the error continues but is actually when I open the parenthesis and not when the second “.” dot is wrote. Withoute the dot, I got an error when trying to run the Flux.train function about broadcasting with σ.(x).

My OS is Linux Mint 19.

6

Really weird. Because it looks more like you’re struggling with finding the sigma symbol rather than anything else I’ll suggest a few things to try.

https://github.com/FluxML/NNlib.jl/blob/master/src/activation.jl
Try using the fn call sigmoid() instead of the symbol?
NNlib I believe is loaded into Flux but you should be able to load it in yourself by adding it.

Make sure you are using a recent version of Julia 1.0+, and that you have a recent version of Flux.

Otherwise, you may need to post more code with explicit error messages. For reference, this function is working fine for me, and has been for a long time.

1 Like