Sameer Agarwal

Change Eigen error handling to more fine grained

1

We usually just check Success, but we should be able to do better by using the other values of the ComputationInfo enum.

and more generally ideas from https://arxiv.org/abs/1201.5885

Questions to add to the documentation

4

This is a tracking issue for people to contribute what questions they would like to see answered in the documentation.

Context: https://groups.google.com/forum/?utm_medium=email&utm_source=footer#!msg/ceres-solver/cPMgCxiMSig/ff9WlJI0CAAJ

It should mention that it should be used with http://ceres-solver.org/nnls_solving.html#_CPPv2N5ceres6Solver7Options28update_state_every_iterationE.

This should work with or without a sparse linear algebra library present, since it only supports the SCHUR_JACOBI preconditioner

Improve documentation on Loss Functions

1

Talk about pitfalls, and how to use them more effectively. For example what happens in a dense problem if you have a single residual versus many residuals. This has come...

Currently the derivatives article uses the Spivak notation and the rest uses the classical d/\nabla/\partial notation and it would be good to make everything read consistently. This likely also implies...

Use ABSL

24

googletest is going to take ABSL as a dependency. Once ABSL gains the ability to do logging, we should replace our use of glog, miniglog and gflags with ABSL. This...