- Binary neural network with 0/1 invert weights.
- Trained with evolutionary reinforcement algorithm, at various cycle counts.
- Swapped memory array is filled with data at inputs and zeros otherwise.
- Output data is OR with any duplicate output data.
- Using int32 storage for memory and weight bits computes 32x input data or weight variations.
Neural OR/NOR node compute [bit1,bit2] -> bit:
Output = (Input1 OR Input2) XOR Weight