a simple keras autoencoder made by chatgpt
pass0.ipynb : setting epsilon = 0 the data array still differs from the perturbed data
foo_2a.py : this code uses https://blog.keras.io/building-autoencoders-in-keras.html and some advice in: https://stackoverflow.com/questions/76752336/inconsistent-output-from-an-auto-encoder
there is a difference between input and predicted data due to the approximation
the randomness upon repeating the same run several times was not resolved
applying : autoencoder to the input data
or
encoder on input data followed by decoder on latent vector
gives the same result which asserts some correctness in the code