We have implemented a genetic algorithm technique which is one of the techniques catering to reinforcement learning on the game called:-  Space Invaders

for results and About refer :- https://bit.ly/2YKmNI4




1) Plot the graphs of the obtained Max score and Avg score in plot.py
2) For playing the game run using loaded_weights.py
3) Two configurations for GA used NN_two_point and NN_uniform
4) weight_data_2.2 = two-point crossover with config 2 of 11 inputs
5) weight_data_2.1 = two-point crossover with config 1 of 21 inputs
6) Assets folder contain images for space invaders game
7) To play the game with a given weight config run the loaded_weight files
    (i) Adjust the input config in main_visual.py
        For weight_data_1.2 - * use config 2 in main_visual.py
                              * Adjust config in feed forward networks use config 2
        For weight_data_1.1 - * use config 1 in main_visual.py
                              * Adjust config in feed forward networks use config 1
        For weight_data_2.1 - * use config 1 in main_visual.py
                              * Adjust config in feed forward networks use config 1
                              * use 23 idx for 124 gen


8)  * We selected a mating pool of best 20 performing weight config as parents
    * Performed two point crossover and uniform cross over technique for mating of rest of 55 children
    * Performed mutation on randomly selected new population

9) Neural Network Config for
    i) 11 input config - 11 x 16 x 3
   ii) 21 input config - 21 x 32 x 3


What is given in 21 input config

Position of nearest enemy on the screen
X coordinates of enemy laser on the screen
Y coordintaes of enemy laser on the screen

What is given in 11 input config

Position of nearest enemy on the screen
X coordinates of enemy laser on the screen