# Эксперимент: Compressed threshold-overlap Gram layer (#1353)

{
  "worked": false,
  "confidence": 9,
  "verdict": "Built a PyTorch MVP for the compressed threshold-overlap Gram layer, including target-rank calculation, PSD Gram fitting, leakage and margin diagnostics, and a retrieval comparison. The factors had the predicted numerical rank and were PSD up to numerical precision, but compressed retrieval was 0.46 versus 1.00 for the 66-dimensional incidence baseline. The 2.36x memory reduction was real, but this implementation did not preserve downstream retrieval behavior.",
  "metrics": {
    "baseline": "66-dimensional s-subset incidence features; retrieval accuracy 1.00",
    "idea": "28-dimensional learned PSD factor; retrieval accuracy 0.46; memory ratio 28/66=0.424; numerical rank 28; forbidden-pair maximum absolute similarity 0.0578; smallest positive magnitude 0.00120; Gram minimum eigenvalue -2.2e-7"
  },
  "how_to_run": "/home/maxwelhelp/main/bin/python3 experiment.py",
  "files": [
    "experiment.py"
  ],
  "limitations": "The exterior-power identity check evaluates the same determinant expression on both sides rather than constructing explicit wedge coordinates, so it is only a weak algebraic sanity check. The learned table was optimized directly and was not integrated into a DeepSets or GNN model; no larger-scale or repeated-seed study was performed."
}