The following notebooks could benefit from figures being added. They can be images or plots generated from within Julia.

• Network Flows
• Geometric Problems

Put semiinteger and semicontinuos contraints to the botton. Same for SOS.

Bring logical conditionals and disjunctions should go up.

#' creates an empty markdown and latex cell in the case of jupyter notebooks which is not needed.

Most of the files there are the same as the scripts. The only difference is that comments are removed.
Comments are not executed anyway... we should just run the script folder.

Hi,
I am opening this issue to sum up a few discussions we had during the JuMP dev calls. I would like to know what are the current plans concerning :

• the automatic generation of the files in notebooks/ in travis
• the status of JuMPTutorials w.r.t. JuMP: do we still plan on merging JuMPTutorials in the JuMP repo?
• also, is there any plan to generate a HTML version of the tutorials, to include them directly in the JuMP's website (e.g. in a cvxpy fashion)?

Again, this issue is just to port the discussion in an asynchronous place, and I would be happy to help on improving JuMPTutorials.

Could you please add a little information to help new Julia users to get set up for the JuMPTutorials.jl in the README?

I installed Julia via Homebrew and installed JuMP and a few solvers via the Julia Pkg manager, and it worked. Then I installed IJulia and for a short time it worked, I could start a Julia notebook from the launcher after jupyter lab.

Then I tried to get the JuMPTutorials.jl to run, and ran this command from your .travis.yml:

julia --project=docs/ -e 'using Pkg; Pkg.instantiate(); Pkg.develop(PackageSpec(path=pwd()))'

Now I'm in deep trouble, I get errors like this when I try to start jupyter lab:

ERROR: LoadError: Failed to precompile IJulia

See https://gist.github.com/cdeil/020d25045b6b4ca4dd67b53426b782b3

I'm experienced with Python and conda, but I'm completely new to Julia, don't know about Julia environments or Project.toml and Manifest.toml or how those interact with my conda environments. Would appreciate a tip how to remove the broken env / install. But my main feature request here is some docs in the README what the "good" or "normal" way to get set up for the Jump tutorials locally is.

Very few changes should be needed beyond replacing with_optimizer.

I don't have the ability to do this since I don't own the repo. @barpit20 you can try going to settings and transferring ownership to JuliaOpt.

What are we doing currently?

Should we run the scripts themselves?

I have seen we have specific files for tests, are they automatically generated?

For the Unit Commitment Model, one of the current constraint reads:

@constraint(uc, [i = 1:2], g[i] <= g_max[i]) #maximum
@constraint(uc, [i = 1:2], g[i] >= g_min[i]) #minimum

it should instead be:

@constraint(uc, [i=1:2], g[i] <= g_max[i] * u[i])
@constraint(uc, [i=1:2], g[i] >= g_min[i] * u[i])

The u variable is the binary unit commitment controller for the optimization. It is declared but not utilized in the example.

like https://github.com/JuliaOpt/Dualization.jl

We need to explore services such as Binder and Google Colab to enable people to try out the tutorials in the browser itself without having to download anything.

Benders Decomposition

We can add links such as:

• one explaining other relevant definitions and the high-level benders algorithm mathematically
• issue/pr related to the status of callbacks

Switching to pandoc from Julia markdown will enable usage of citations in Weave 0.4.1

It would be nice if there was a CI check that the notebooks correspond to the script files that are checked in. It could be easy to forget to regenerate a notebook when modifying a script.

We can possibly try to tabulate the data in the power systems tutorial by storing it in a data frame.

As mentioned in the style guide, use the plural form wherever there are 3 or more consecutive lines of @variable or @constraint. Modelling notebooks are where this change will mostly be required.

As of Weave 0.5.1. there is a new notebook method to convert Weave documents to Jupyter notebooks using nbconvert. It runs all the code blocks in the notebook which is better for us.

Detail a step-by-step procedure.
Explain which file should be written (point to and example)
Explain which file ares generated automatically (and possible caveats)

Just for tracking.

The failure is only on Julia 1.5 at this point, and seems to come from the Weave module.
The error message is

ERROR: LoadError: InitError: Evaluation into the closed module `Markdown` breaks incremental compilation because the side effects will not be permanent. This is likely due to some other module mutating `Markdown` with `eval` during precompilation - don't do this.

Is it possible to generate files with travis or something similar?

https://github.com/JuliaOpt/JuMPTutorials.jl/blob/5f11c956c6184fcc5eb3825ede3998df6c3e48bd/script/modelling/experiment_design.jl#L59

"eye" is a Matlab-ism that is discouraged in Julia. @barpit20 Can you try putting I directly into the @constraint? It it doesn't work, please file another feature request on JuMP (or add to the Diagonal feature request).

S1 and S2 are not commonly used terminology for Special Ordered Sets and hence we should stick to SOS1 and SOS2.

Very cool! This is going to be a great resource.

Just a query re the tests. It seems like the tests run all of the code in the notebooks, but I couldn't find any checks using @test that the output is what we expect. Is there a way to add some tests? Or is it sufficient to just run the code and check that it doesn't error?

Hello,

I have created a tutorial for the N-Queens problem that I'd be more than happy to contribute here if you all are interested. It can be viewed here:

https://github.com/mthelm85/nqueens/blob/master/N-Queens.ipynb

I'm happy to flesh it out, make changes, etc. If you all would like to include this, let me know and I'll figure out how to submit a pull request ; )

Add semicolons at the end of code blocks where we want to suppress output

I was wondering, what is the difference between https://github.com/JunoLab/Weave.jl and https://github.com/fredrikekre/Literate.jl and what makes Weave more appropriate for this project ?

Currently in nonlinear_modelling.ipynb for the MLE example I see this result:

μ = 0.0
mean(data) = -0.02534297022835935
σ^2 = 1.0
var(data) = 1.0112373171501061
MLE objective: -0.0

Is mu=0 and objective=0 a bug? Shouldn't mu match mean?

I have IJulia install issues, can't get the notebook to run unfortunately to investigate.

I have used a few different ways to write the models in the math mode i.e. spacing differs between some notebooks. This isn't really high priority though.

1. Introduction

• An Introduction to Julia
• Getting started with JuMP
• Variables, Constraints and Objective
• Solutions
• Solvers

2. Using JuMP

• Reading from Data files
• Problem Modification
• Nonlinear Modeling

3. Optimization Concepts

• Integer Programming
• Conic Programming
• Benders Decomposition

4. Modeling Examples (Currently a list of topics to pick examples from)

• Sudoku
• Power Systems
• Network Flows
• Statistics
• Approximation and Fitting
• Geometric Problems
• Control theory
• Graph Theory
• Economics/Econometrics
• Finance
• Rocket Control
• Travelling Salesman Problem

Note that this is a tentative list and some notebooks can be combined or split into two parts. Topics such as non-linear modelling, callbacks and column generation are ongoing dev efforts. Things are subject to change there and notebooks for them will only be created after they are finalized.

The current function only weaves a file in case it detects a change. Other functions such as weaving without a check would be useful.

In the Jupyter notebook renderings at least, the notebook title is the same size as the subsection titles. It would be nice if the notebook title was more clearly distinguished by size. Can the section and subsection titles be made smaller?

There is also some repetitiveness between section titles, and in some cases I think some notebooks have more sections/subsections than is needed. eg in https://nbviewer.jupyter.org/github/barpit20/JuMPTutorials.jl/blob/master/notebook/modelling/experiment_design.ipynb

in the experiment design notebook, for example, we call rand. should we just hardcode a small V matrix? or set Random.rseed(123) say? for reproducibility.

We have 1.0 in the entire julia intro.

There is some error while running xelatex but if the file is weaved directly there is no such error.

https://nbviewer.jupyter.org/github/JuliaOpt/JuMPTutorials.jl/blob/master/notebook/optimization_concepts/benders_decomposition.ipynb

┌ Info: Precompiling JuMP [4076af6c-e467-56ae-b986-b466b2749572]
└ @ Base loading.jl:1192
ERROR: LoadError: ArgumentError: Package MutableArithmetics [d8a4904e-b15c-11e9-3269-09a3773c0cb0] is required but does not seem to be installed:
 - Run `Pkg.instantiate()` to install all recorded dependencies.

Stacktrace:
 [1] _require(::Base.PkgId) at ./loading.jl:929
 [2] require(::Base.PkgId) at ./loading.jl:858
 [3] require(::Module, ::Symbol) at ./loading.jl:853
 [4] include at ./boot.jl:317 [inlined]
 [5] include_relative(::Module, ::String) at ./loading.jl:1044
 [6] include(::Module, ::String) at ./sysimg.jl:29
 [7] top-level scope at none:2
 [8] eval at ./boot.jl:319 [inlined]
 [9] eval(::Expr) at ./client.jl:393
 [10] top-level scope at ./none:3
in expression starting at /home/mbesancon/.julia/packages/JuMP/CZ8vV/src/JuMP.jl:16

First thank a lot for putting these examples together!

The entropy maximization example under "Conic Programming" contains a bug. The correct exponential cone for this example is
(t,x,1)∈Kexp
instead of
(1,x,t)∈Kexp .

Hence the constraint should be

@constraint(model, con[i = 1:n], [t[i]], x[i], 1] in MOI.ExponentialCone())

When using Weave to convert a tutorial which reads from another data file, Weave uses the path of the current directory instead of the path of the file being weaved. This throws an error that the file cannot be found.

Other packages (JuMP, MOI, Dualization etc) don't commit Manifest.toml, only Project.toml.
You can see the .gitignore of these packages.

There seems to be a latex error but I have not used any latex for this notebook. WIll have to investigate.

see http://www.juliaopt.org/JuMP.jl/dev/style/
there are some subtle points like whitespace around binary operators, function and variable naming, etc that may require some minor changes to the code already committed

since this isn't really a Julia "package", but rather a repo for tutorials, would it make sense to reduce clutter and remove src and test folders and other unnecessary files that were autogenerated?

I think we should keep tutorials accessible from all the three systems.

The notebooks in the old juliaopt-notebooks repo had the Creative Commons Attribution-ShareAlike 4.0 International License. Do we need to look into licensing for these tutorials as well?

@mlubin @juan-pablo-vielma