| license: mit | |
| tags: | |
| - pytorch | |
| - safetensors | |
| - threshold-logic | |
| - neuromorphic | |
| # threshold-absolutevalue4 | |
| Absolute value of 4-bit 2's complement signed integer. | |
| ## Function | |
| abs4(a) = |a| where a is interpreted as 4-bit 2's complement (-8 to +7) | |
| ## Truth Table | |
| | Input | Signed | |Abs| | Output | | |
| |-------|--------|------|--------| | |
| | 0000 | +0 | 0 | 0000 | | |
| | 0001 | +1 | 1 | 0001 | | |
| | 0010 | +2 | 2 | 0010 | | |
| | 0011 | +3 | 3 | 0011 | | |
| | 0100 | +4 | 4 | 0100 | | |
| | 0101 | +5 | 5 | 0101 | | |
| | 0110 | +6 | 6 | 0110 | | |
| | 0111 | +7 | 7 | 0111 | | |
| | 1000 | -8 | 8 | 1000 | | |
| | 1001 | -7 | 7 | 0111 | | |
| | 1010 | -6 | 6 | 0110 | | |
| | 1011 | -5 | 5 | 0101 | | |
| | 1100 | -4 | 4 | 0100 | | |
| | 1101 | -3 | 3 | 0011 | | |
| | 1110 | -2 | 2 | 0010 | | |
| | 1111 | -1 | 1 | 0001 | | |
| ## Architecture | |
| 5-layer circuit implementing conditional 2's complement negation: | |
| - For positive (a3=0): output = input | |
| - For negative (a3=1): output = ~input + 1 | |
| Key formulas for negative path: | |
| - o0 = a0 (always) | |
| - o1 = ~a1 XOR ~a0 | |
| - o2 = ~a2 XOR (~a1 AND ~a0) | |
| - o3 = 1 only for input 1000 (-8) | |
| ## Parameters | |
| | | | | |
| |---|---| | |
| | Inputs | 4 | | |
| | Outputs | 4 | | |
| | Neurons | 23 | | |
| | Layers | 5 | | |
| | Parameters | 145 | | |
| | Magnitude | 64 | | |
| ## License | |
| MIT | |