Sensor Input
Draw Digit
Comparator Pipeline
Input
→
XOR Distances
→
-
Sum(LUT)
Neural Core Control
Strict MNIST Mode. Local files required.
Activation Output
Prediction
-
System Offline
Core Logic: Comparator LUTs
1. 128 Parallel Banks (Neurons) The image is converted into 128 integers (32 bits each) using randomized rank comparisons.
2. Hardware XOR Comparator
For each bank, we compute the Hamming Distance to a fixed random mask:
dist = popcount(Input_Bank ^ Mask_Bank)
3. Lookup Table (LUT) Summation
The distance acts as an index into a learned Lookup Table A. The final score is the sum of these looked-up values across all 128 banks.
Score = Σ A[dist]
> System initialized. Ready to initialize.