Generating Test Data for our Neural Net

Our upcoming LED-digit neural net recognizer will need sample data for training and testing. All of you will generate this data.

There are 128 possible combinations of the (simulated) LEDs. Only ten of these represent digits. But humans also recognize near-correct patterns, so our neural net should too. For each of the 128 patterns, you will register which of the digits it most looks like. You can also register that a pattern looks doesn't look like a digit.

We will use all of the students' data as input. Presumably, you'll all agree on the ten correct digits. On the near-correct patterns, there will likely be some disagreement on what the appropriate answer is. But, we will train the neural net even on such conflicting data.

Run this code, and generate data. When you are done, email the resulting printed data to comp200@.