In your version of flappy bird, the score is assigned based on the distance traveled. Hence, there's no way a bird has score = 0 in a generation. But, in my version, I have assigned points based on the number of pipes crossed. And there's a good chance none of the birds make it in the first generation. Hence, the total score of the generation = 0;

https://github.com/shiffman/NOC-S18/blob/95b801d6e0cefeb9b8d2ac1997d0b68bb46d95d7/week10/neuroevolution-flappybird/ga.js#L75-L79

In this algorithm, since r remains constant the loop runs forever until the index > birds.

https://github.com/shiffman/NOC-S18/blob/95b801d6e0cefeb9b8d2ac1997d0b68bb46d95d7/week10/neuroevolution-flappybird/ga.js#L56-L59

Likewise while calculating the fitness, the scores will be divided by zero resulting in NaN fitness for all birds

How can these issues be solved without changing the scoring system ?

I'tried to use the mutate function and that appeared :
TypeError: func is not a function matrix.js:102:23

#issue detected in week10 flappy bird

run best button

i was attempting to run this on my phone. And when i clicked on the run best button it launches the best neural network yet and then it launches the training immediately right after that.

high score

when running the "best so far" the high score never resets itself even after the agent lost the game