es hyperneat support

[ai-nick]

i have added the supporting to code to use tensorflow nets for es hyperneat, unlike many other implementations that only support 2d substrates you can us any dimension of substrate for with this algorithm allowing you to utilize very large search spaces in task domains that require this.

[codecov-io]

Codecov Report

Merging #4 into master will decrease coverage by 12.89%. The diff coverage is 0%.

@@            Coverage Diff             @@
##           master       #4      +/-   ##
==========================================
- Coverage   88.61%   75.71%   -12.9%     
==========================================
  Files          22       24       +2     
  Lines        1923     2248     +325     
==========================================
- Hits         1704     1702       -2     
- Misses        219      546     +327

Impacted Files	Coverage Δ
tf_neat/substrate.py	0% <0%> (ø)
tf_neat/es_hyperneat.py	0% <0%> (ø)
tf_neat/maze.py	90.9% <0%> (-2.28%)	:arrow_down:

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update 64d9400...8d517ee. Read the comment docs.

[crisbodnar]

@ai-nick thanks for contributing! Can you write some tests to start with? I will have a look over your PR in the meantime.

[ai-nick]

yeah i actually made a test experiment and i must have something off because i got an error on activation, ill get something pushed up with the experiment and the fix

[crisbodnar]

Can you add the tests to the PR? Check https://github.com/crisbodnar/TensorFlow-NEAT/tree/master/tests for some example tests for the existent models. I will have a look once the tests are added and they pass.

[ai-nick]

oh right i see those now, yeah ill right some for my ESNetwork class

[ai-nick]

hey been rather busy but have not forgot about this and will be revisiting over the weekend

[ai-nick]

i might close this for the time being, still working out some kinks with eshyperneat, and when i finish that i have one more algo (of my own creation) that i may include in the pr if you dont mind