##Evolving an Inverted Pendulum Controller with a Genetic Algorithm
See here for more details.
###Running the demo
A working set of weights and biases are provided in weights_working.txt
and
biases_working.txt
. This is used by the provided Mathematica demo
inverted pendulum neural network.nb
(evaluate all cells after opening it).
###Running the genetic algorithm Python 2.7 and numpy is required. Run
python trainer.py
to start the training. Computation is automatically parallelized across all
available cores. The training parameters are set at the top of trainer.py
.