michaelhush/M-LOOP

When optimizing one parameter, there were some issues reimporting the
saved files for the visualizations to work. This was due to the
problematic corner case of zero D or one D with one element arrays in
numpy. This has now been sanitized. Also fixed some critical typos in
the documentation.

Fixed a bug where an attribute wasn’t present in the learner class. Was
a problem when attempting to plot the visualizations from a file.

Added support for previous data files to be imported into a gaussian
process learner.

…isn't set

at the end. This was causing an exception to be thrown when trying to
get costs from the queue.

When generation_num=1, if the new_params_event is set first then the
learner will try to get the cost when the queue is empty, causing an
exception.

…to NeuralNetA

Conflicts:
	mloop/controllers.py
	mloop/learners.py

We need to specify nnlearner as a package. More subtly, because of TF
we can only run NNI in the same process in which it's created. This
means we need to wait until the run() method of the learner is called
before constructing the impl.

I've pulled the actual network logic out into a new class, to
keep the TF stuff separate from everything else and to keep a
clear separation between what's modelling the landscape and
what's doing prediction.

There appears to be some issues with multiprocessing and gaussian
process but only on MacOS, and possibly just my machine. So I’ve
removed all the testing statements I had in the previous commit.

Branch should be ready now to integrate in a genuine NN.

Visualizations now work for NN and GP learners.

Mysterious bug has appeared in GP. The scikit-learn stops providing
uncertainty predictions after being fit for a certain number of times.

Commiting so I can change branch and investigate.

Git complains to me about them when I touch nearby lines, so I figured
it was easier just to fix them.

Added a controller and learner for the neural net.

Also added a new class MachineLearnerController which GaussianProcess
and NeuralNet both inherit from.

I broke the visualizations for GPs in this update. But all the tests
work.

Added some updates to docstrings and test unit parameters.

Updated the documentation.
Candidate for new version to be released on PyPI

Previously the training runs had to be completed before M-LOOP would
halt. This lead to unintuitive behavior when the halting conditions
were early on in the optimization process.

M-LOOP now halts immediately when any of the halting conditions are
met.

Fixed bug with GP fitting data with bad runs.

Candidate for version 2.1.0 release.

Update to documentation complete. Now describes how to use the shell
interface, the differential evolution optimizer and using M-LOOP as an
MPI.

Documentation has been updated to explain all the added features and
also how to use M-LOOP as a python API. Still needs proof reading.