update tutorial page

doc/README.md

@@ -9,11 +9,10 @@ How to make a release
 - create a branch on the master, called release_<version>, make release edits there
 - increase version number in iminuit/version.py
 - update doc/changelog.rst
-- check that new tutorials are listed in the tutorials section of the docs
 - run `make integration` to do integration tests (if these fail, add tests to iminuit!)
 - run `.ci/download_azure_artifacts.py` to download all wheels from the latest pipeline
 - run `python -m twine upload dist/*` if everything looks ok
   (missing files can be uploaded later, but existing files cannot be overridden!)
-- merge release branch to master (do not squash!)
+- merge release branch to master
 - create release on Github
 - conda-forge should pick this up automatically and generate conda packages

doc/tutorials.rst

@@ -5,45 +5,6 @@
 Tutorials
 =========
 
-All the tutorials are in tutorial directory. You can view them online, too.
+All the tutorials are in tutorial directory of the iminuit repository and `can be viewed online <http://nbviewer.ipython.org/urls/raw.github.com/scikit-hep/iminuit/master/tutorial/>`_.
 
-`Basic tutorial <http://nbviewer.ipython.org/urls/raw.github.com/scikit-hep/iminuit/master/tutorial/basic_tutorial.ipynb>`_
----------------------------------------------------------------------------------------------------------------------------------
-
-Covers the basics of using iminuit.
-
-`iminuit and automatic differentiation with JAX <http://nbviewer.ipython.org/urls/raw.github.com/scikit-hep/iminuit/master/tutorial/automatic_differentiation.ipynb>`_
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------
-
-How to compute function gradients for iminuit with jax_ and accelerate Python code with JAX's JIT compiler. Spoiler: a **32x** speed up over plain numpy is achieved. Also discusses how to do a least-squares fit with data that has uncertainties in *x* and *y*.
-
-`How to accelerate and parallelize cost functions with Numba <http://nbviewer.ipython.org/urls/raw.github.com/scikit-hep/iminuit/master/tutorial/jit_compilation_andparallelization_with_numba.ipynb>`_
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
-
-Shows how cost functions can be accelerated by Just-In-Time compilation with numba, and how important it is to optimize the most time-consuming hot spot.
-
-`iminuit and an external minimizer <http://nbviewer.ipython.org/urls/raw.github.com/scikit-hep/iminuit/master/tutorial/iminuit_and_external_minimizer.ipynb>`_
---------------------------------------------------------------------------------------------------------------------------------------------------------------
-
-iminuit can run the HESSE algorithm on any point of the cost function. This means one can effectively combine iminuit with other minimizers: let the other minimizer find the minimum and only run iminuit to compute the parameter uncertainties. This does not work with MINOS, which requires that MIGRAD is run first.
-
-`How to write a generic least-squares cost function <http://nbviewer.ipython.org/urls/raw.github.com/scikit-hep/iminuit/master/tutorial/generic_least_squares_function.ipynb>`_
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
-
-Explains how iminuit reads function signatures and how the tooling can be used in a generic least-squares cost function that forwards the signature of the fitted model to iminuit.
-
-`Uncertainty computation in iminuit <http://nbviewer.ipython.org/urls/raw.github.com/scikit-hep/iminuit/master/tutorial/hesse_and_minos.ipynb>`_
-------------------------------------------------------------------------------------------------------------------------------------------------
-
-This is less of a tutorial and more of a write-up on how MINUIT (and thus iminuit) compute uncertainties from likelihood functions.
-
-Outdated Cython tutorials
--------------------------
-
-The following two tutorials are outdated. Users who want to speed up their fits should try the just-in-time compilers provided by numba_ or jax_ in CPython or use iminuit in PyPy to accelerate the computation. This is much simpler than using Cython and may achieve even better performance.
-
-- `Advanced tutorial <http://nbviewer.ipython.org/urls/raw.github.com/scikit-hep/iminuit/master/tutorial/advanced_tutorial.ipynb>`_.
-  Shows how to speed up the computation of the cost function with Cython.
-
-- `Hard Core Cython tutorial <http://nbviewer.ipython.org/urls/raw.github.com/scikit-hep/iminuit/master/tutorial/hard_core_tutorial.ipynb>`_.
-  Goes into more detail on how to use Cython.
+It is recommended to start with the basic tutorial. All other tutorials are optional, for those who want to know more about specific aspects.