--- license: mit tags: - formal-verification - coq - threshold-logic - neuromorphic --- # tiny-7OutOf8-verified Formally verified 7-out-of-8 threshold gate. Single threshold neuron with 100% accuracy. ## Architecture | Component | Value | |-----------|-------| | Inputs | 8 | | Outputs | 1 | | Neurons | 1 | | Parameters | 9 | | Weights | [1, 1, 1, 1, 1, 1, 1, 1] | | Bias | -7 | | Activation | Heaviside step | ## Key Properties - 100% accuracy (256/256 inputs correct) - Coq-proven correctness - Single threshold neuron - Integer weights - Fires when at least 7 of 8 inputs are true ## Usage ```python import torch from safetensors.torch import load_file weights = load_file('sevenoutof8.safetensors') def sevenoutof8_gate(bits): # bits: list of 8 binary values inputs = torch.tensor([float(b) for b in bits]) weighted_sum = (inputs * weights['weight']).sum() + weights['bias'] return int(weighted_sum >= 0) # Test print({func}_gate([0,0,0,0,0,0,0,0])) # 0 (0/7, below threshold) print({func}_gate([1,1,1,1,1,1,0,0])) # 0 (6/8, below threshold) print({func}_gate([1,1,1,1,1,1,1,0])) # 1 (7/8, at threshold) print(sevenoutof8_gate([1,1,1,1,1,1,1,1])) # 1 (8/8, above threshold) ``` ## Verification **Coq Theorem**: ```coq Theorem sevenout_eight_correct : forall x0 x1 x2 x3 x4 x5 x6 x7, sevenout_eight_circuit [x0; x1; x2; x3; x4; x5; x6; x7] = sevenout_eight_spec [x0; x1; x2; x3; x4; x5; x6; x7]. ``` Proven axiom-free via: 1. **Exhaustive**: All 256 inputs verified 2. **Universal**: Quantified proof over boolean combinations 3. **Algebraic**: Hamming weight ≥ 7 Full proof: [coq-circuits/Threshold/SevenOutOfEight.v](https://github.com/CharlesCNorton/coq-circuits/blob/main/coq/Threshold/SevenOutOfEight.v) ## Circuit Operation Input with h true bits (Hamming weight h): - Weighted sum: h - 7 - Output: 1 if h ≥ 7, else 0 ## Citation ```bibtex @software{tiny_7outof8_prover_2025, title={tiny-7OutOf8-verified: Formally Verified 7-out-of-8 threshold gate}, author={Norton, Charles}, url={https://huggingface.co/phanerozoic/tiny-7OutOf8-verified}, year={2025} } ```