Add benign overfitting theory, double descent, LOOCV Ridge tuning, condition number diagnostics (2026-04-25)

Add lstsq conv research: Ridge regularization, stride_tricks, benign overfitting theory (2026-04-25)

Create simple-onnx-processing-tools-README.md

Replace TODO with roadmap (2026-04-25)

Add deep competitive analysis from notebook dissection (2026-04-25)

Update LEARNING.md: PyTorch conv results, lstsq overfitting analysis

v4.2: Add PyTorch learned conv solver (single+two-layer, multi-seed, ternary snap). Needs GPU to be practical - use on Kaggle with --conv_budget 60

Add LEARNING.md - full history, mistakes log, competitive intel, technical deep-dives

Split SKILL.md (rules/quick-ref) + LEARNING.md (history/mistakes/analysis)

v4.1: color_map via channel Gather for permutation maps (+15 pts), fix non-permutation fallback to Conv 1x1

v4: ARC-GEN validation + fitting, new analytical solvers (shift/mirror/crop/quad), s_flip opset fix, static profiler, submission.csv

v4: ARC-GEN validation, new analytical solvers, static profiler, s_flip opset fix

Add SKILL.md - complete knowledge base for NeuroGolf solver

Create ARC-GEN-README.md

Create skill-creator.md

Update README.md

Upload neurogolf-2026-solver-notebooks.zip

Upload ARC-GEN-100K.zip

Update neurogolf_solver.py

Create PyTorch -to- ONNX.md

Update TODO.md

Create TODO.md

Update README for v3 (306/400)

v3: 306/400 - Fix GatherElements->Gather (opset 10), add concat_enhanced, varshape_spatial_gather, conv_var_diff solvers

Update README.md

Update neurogolf_solver.py

Update neurogolf_solver.py

Create neurogolf_utils.py

v2: 293/400 tasks solved (was 128). Added variable-shape conv and diff-shape conv solvers.

Update README with full instructions

v2: add --device cpu/cuda/auto flag, fix providers

Add README with run instructions

Upload complete neurogolf solver v1

Upload solver script

Upload neurogolf_solver.py

initial commit