What Neural Network to use for AI Mouse Movement

Question

I am trying to make a fuction that takes in 2 (x,y) coordinates and returns and array of coordinates for where the mouse should be every 0.05 second (or about that)...

(The whole aim of the program is to have random/not smooth mouse movement to mimic a human's mouse movement)

EG INPUTS : [(800,600), (300,400)] OUTPUTS : [(800,600),(780,580),...,(300,400)]

I would like to know which type of neural network I should use to get the correct outputs.

I'm kind of new to the subject of neural nets but I have a decent understanding and will research the suggestions given.

A pointer in the right direction and some links would be greatly appreciated!

Answer 1

Using a neural network for this task seems like total over kill to me. It seems like you have 2 inputs, each which have an X and a Y coordinate with one representing the initial position and one representing the final position of the mouse.

There are a ton of ways you can introduce randomness into this path in hard to detect ways that are much simpler than a neural network. Use some weird random number generator with bizarre personal logic in if statements to determine the amounts to add within some range to the current value on each iteration. You could use a neural net, but again i think it's overkill.

As far as what type of neural net you need to use, i would just start with an out of the box one from a tutorial online (tensorflow and sklearn are what i've used) and tweak the hyper parameters to see what makes the model better.

Answer 2

If you're trying to predict where the mouse should be based on the position of something else, a simple ANN will do the job.

Are you trying to automate tasks, like have a script that controls a game? A Recurrent Neural Network like an LSTM or GRU will take history into account.

I know you're doing this as a learning exercise, but if you're just trying to smooth mouse movement a simple interpolation algorithm might work.