Rohit Ghumare f9f044a07c Fix all 41 CodeRabbit review findings ...

Code correctness:
- Fix Layer/MLP output unwrapping in autodiff to prevent TypeError
- Use weighted degree in graph Laplacian, handle dangling nodes in PageRank
- Fix isolation forest c(n) base case and score normalization
- Fix test label leakage in feature selection and threshold tuning
- Remove MSE censoring that biased learning curve averages
- Handle edge case of single minority sample in SMOTE
- Build random baseline from training labels, not test labels
- Re-normalize stationary distribution after clipping numerical errors
- Add input validation (spectral clustering, MH sampler, NB counts,
  gamma shape, forecast history, convolution lengths)

Documentation:
- Add __radd__ to autodiff docs for reflected arithmetic
- Correct GCN formula to include self-loops (A_hat = A + I)
- Soften NB probability calibration claim
- Fix Optuna pruning example to use trial.report/should_prune
- Tune threshold on validation data not test data in docs
- Qualify Bayesian early stopping claim
- Fix comparison snippet in hyperparameter tuning docs
- Add language IDs to fenced code blocks (MD040)
- Escape pipe in table cell, fix step references
- Soften absolute metric claims, add target encoding leakage note
- Use endpoint-free time grid for FFT alignment

Style:
- Remove unnecessary f-string prefixes across 4 files

2026-03-20 22:02:18 +05:30

..

skill-statistical-testing.md

Fix all 41 CodeRabbit review findings

2026-03-20 22:02:18 +05:30