|
|
--- |
|
|
license: mit |
|
|
tags: |
|
|
- pytorch |
|
|
- safetensors |
|
|
- threshold-logic |
|
|
- neuromorphic |
|
|
- encoder |
|
|
--- |
|
|
|
|
|
# threshold-priorityencoder8 |
|
|
|
|
|
8-to-3 priority encoder. Outputs 3-bit binary encoding of highest-priority active input. |
|
|
|
|
|
## Function |
|
|
|
|
|
priority_encode(i7..i0) -> (y2, y1, y0, valid) |
|
|
|
|
|
- i7 = highest priority, i0 = lowest priority |
|
|
- y2,y1,y0 = 3-bit binary encoding of highest active input index |
|
|
- valid = 1 if any input is active |
|
|
|
|
|
## Architecture |
|
|
|
|
|
**Layer 1: 8 neurons (h7..h0)** |
|
|
|
|
|
Each hk detects "ik is the highest active input": |
|
|
- hk fires when ik=1 AND all higher-priority inputs are 0 |
|
|
- h7: weights [1,0,0,0,0,0,0,0], bias -1 |
|
|
- h6: weights [-1,1,0,0,0,0,0,0], bias -1 |
|
|
- ... |
|
|
- h0: weights [-1,-1,-1,-1,-1,-1,-1,1], bias -1 |
|
|
|
|
|
**Layer 2: 4 neurons** |
|
|
|
|
|
- y2 = h7 OR h6 OR h5 OR h4 |
|
|
- y1 = h7 OR h6 OR h3 OR h2 |
|
|
- y0 = h7 OR h5 OR h3 OR h1 |
|
|
- v = any h active |
|
|
|
|
|
## Parameters |
|
|
|
|
|
| | | |
|
|
|---|---| |
|
|
| Inputs | 8 | |
|
|
| Outputs | 4 | |
|
|
| Neurons | 12 | |
|
|
| Layers | 2 | |
|
|
| Parameters | 108 | |
|
|
| Magnitude | 68 | |
|
|
|
|
|
## Usage |
|
|
|
|
|
```python |
|
|
from safetensors.torch import load_file |
|
|
import torch |
|
|
|
|
|
w = load_file('model.safetensors') |
|
|
|
|
|
# (see model.py for full implementation) |
|
|
|
|
|
# Example: i5 is highest active (index 5 = 101) |
|
|
# priority_encode(0,0,1,0,0,0,0,0, w) -> (1, 0, 1, 1) |
|
|
``` |
|
|
|
|
|
## License |
|
|
|
|
|
MIT |
|
|
|
