Эксперимент: Cut-Aware Augmentation Filtering (#1370)

{ "worked": true, "confidence": 8, "verdict": "Built a self-contained cut-aware augmentation graph filter with cut-ratio computation, exponential policy weighting, graph regularization, and fixed-seed synthetic evaluation. The math check confirmed scale invariance and strongly favored the low-cut graph; across three seeds, cut-aware weighting reduced predicted cut mass from 0.2366 to 0.0092 and improved accuracy from 0.9802 to 0.9975. This is a clear toy-benchmark signal, not yet evidence of CIFAR-scale effectiveness.", "metrics": { "baseline": "Uniform policy mixing: mean accuracy 0.98025 +/- 0.00175; mean predicted cut ratio 0.23663 +/- 0.00581.", "idea": "Cut-aware weighting: mean accuracy 0.99753 +/- 0.00349; mean predicted cut ratio 0.00923 +/- 0.00303; harmful shuffled-policy weight approximately 0.00018-0.00020." }, "how_to_run": "/home/maxwelhelp/main/bin/python3 experiment.py && /home/maxwelhelp/main/bin/python3 robustness.py", "files": [ "experiment.py", "robustness.py", "results.json", "robustness.json" ], "limitations": "Tested only a three-class 2D Gaussian toy dataset with feature-built graphs and ground-truth labels for policy scoring. Confidence-threshold pseudo-labeling, EMA estimates, q_min/warmup, random policy dropping, real image augmentations, CIFAR/STL-10, sparse graph implementation, GPU behavior, and matched FLOP/runtime measurements were not tested." }