In GaussianProcess.jl it is throwing error if I choose `MeanConst()`

Specific Domains Statistics
1

Hi

In the following code, I am trying to implement a constant mean function from GaussianProcesses.jl. But it is throwing an error stating no method matching MeanConst()

using GaussianProcesses
n = 100;
x = 2π * rand(n);
y = sin.(x) + 0.2*randn(n);
#mZero = MeanZero()
mConst = MeanConst()
kern = SE(0.0,0.0) ## squared exponential kernel
logObsNoise = -1.0
gp = GP(x, y, mConst, kern, logObsNoise)
optimize!(gp)
plot(gp)

It is throwing the following error.

MethodError: no method matching MeanConst()
Closest candidates are:
  MeanConst(::Float64) at ~/.julia/packages/GaussianProcesses/dOato/src/means/mConst.jl:23

Stacktrace:
 [1] top-level scope
   @ In[137]:5
2

As suggested in the error, the only method is MeanConst(::Float64), meaning you need to provide a constant as an argument - there is no default method with zero arguments.

1 Like
3

Thank you. This makes sense. I provided a value and it worked. Thanks again.

4

Dear Daniel

Thanks again for pointing out the correct way to feed the arguments. One more question, how can we feed the MeanLin function?

https://stor-i.github.io/GaussianProcesses.jl/latest/mean/#GaussianProcesses.MeanLin

Thanking you in advance.
Best regards,
Sourish

5

Judging from the source here GaussianProcesses.jl/mLin.jl at 1bcdf2dd8a988b91b0661e7ad1504d5fa66d9e65 · STOR-i/GaussianProcesses.jl · GitHub (bottom right button of that documentation), you just pass it a vector of coefficients.

6

Yes - that is what I tried. But facing some errors. Thanks again. I will figure it out. Thanks again.

Best regards,
Sourish

7

What was the error?

julia> MeanLin([1.0,2.0,3.0])
Type: MeanLin, Params: [1.0, 2.0, 3.0]
8

Dear Daniel:

Thanks again for your generous help. I tried your code:

MeanLin([1.0,2.0,3.0])

and I am getting this error.

MethodError: no method matching MeanLin(::Vector{Float64})
Closest candidates are:
  MeanLin() at In[161]:1

Do I need to update my system?

9

You might’ve accidentally re-defined the function, e.g. maybe something you did

MeanLin() = ...

or similar. You can see this from In[161]:1 being the location of the method, when really it should be somewhere in GaussianProcesses.jl. Once you restart, it’ll be fixed

10

Oh dear - yes possible. Let me restart my session.

Sourish