Implement a program that's a multi-layer feedforward back-propagation algorithm (or in short, an artificial neural network, NOT a perceptron!) to recognize a 2 by 2 block (like the one below) to decide if it is "bright" or "dark". It needs 16 samples for training inputs. It can be written in C++, Java, or Python. If in Python, please take a screenshot so I can apply appropriate indentation. Thank you!

Please don't just take code from some other website. I can google to tell, and other people can too.

To be classified as "bright", it must have 2, 3, or 4 white pixels. For "dark", it must be 0 or 1 pixels.

Here are 16 training inputs:

[-1,-1,-1,-1]

[-1,-1,-1,1]

[-1,-1,1,-1]

[-1,-1,1,1]

[-1,1,-1,-1]

[-1,1,-1,1]

[-1,1,1,-1]

[-1,1,1,1]

[1,-1,-1,-1]

[1,-1,-1,1]

[1,-1,1,-1]

[1,-1,1,1]

[1,1,-1,-1]

[1,1,-1,1]

[1,1,1,-1]

[1,1,1,1]