Linear regression using flux.jl in julia

0
votes

I am looking for an alternative to MLJLinearModel for linear regression in Julia. I have read a few blogs describing flux.jl as a better and powerful package to perform ML tasks. However, I have not found a well-detailed tutorial directed specifically toward linear regression. Please do recommend either an approach to perform this task or sources which can help me understand the process in details.

Thank you in advance!

1
julialinear-regression

1 Answers

2
votes

I recommend using GLM for this job. For an example:

using GLM, DataFrames

a, b = 2, 3;
dat = DataFrame(x=1:1000, y=a .* (1:1000) .+ b .+ randn(1000))

ols = lm(@formula(y ~ x), dat)

which yields:

StatsModels.TableRegressionModel{LinearModel{GLM.LmResp{Vector{Float64}}, GLM.DensePredChol{Float64, LinearAlgebra.Cholesky{Float64, Matrix{Float64}}}}, Matrix{Float64}}

y ~ 1 + x

Coefficients:
───────────────────────────────────────────────────────────────────────────
               Coef.   Std. Error         t  Pr(>|t|)  Lower 95%  Upper 95%
───────────────────────────────────────────────────────────────────────────
(Intercept)  3.0607   0.0641567       47.71    <1e-99    2.9348     3.1866
x            1.99989  0.000111039  18010.65    <1e-99    1.99967    2.00011
───────────────────────────────────────────────────────────────────────────

You can always just get the coefficients:

julia> coef(ols)
2-element Vector{Float64}:
 3.0607023932922464
 1.9998906641774181