Fix DiceGenetic.mate continuous-mutation typo pinning to lower bound (#441)

[jbbqqf]

Summary

DiceGenetic.mate() mutates a continuous feature with
np.random.uniform(low, high), but the call passed
feature_range[feat_name][0] for both arguments. As a result every
continuous mutation collapsed to the lower bound of the feature's
declared range, leaving the genetic search to explore continuous axes
only via the parent-gene branches (prob < 0.80). Half of the
diversity-injection strategy was a no-op.

The two adjacent np.random.uniform calls in the same file
(do_random_init at line ~92 and do_KD_init at line ~124) both
correctly pass low, high, so this is a clear typo rather than an
intentional design choice — and it matches several user reports that
the genetic explainer "ignores" the user's permitted_range /
features_to_vary intent on continuous columns (see #441 and #260
comment threads).

Why

One-character typo fix in the else: mutation branch of
DiceGenetic.mate. Restores the documented diversity behaviour of the
genetic algorithm.

I left a 4-line comment at the call site explaining the prior bug —
without it, a future reviewer or refactor could easily re-introduce
the same typo in another mutation path.

Reproduce BEFORE/AFTER yourself (copy-paste)

set -e
cd /tmp && rm -rf DiCE-441 && git clone https://github.com/interpretml/DiCE.git DiCE-441
cd DiCE-441
pip install -q -e . pytest

git fetch https://github.com/jbbqqf/DiCE.git fix/441-genetic-mate-uniform-typo
git checkout FETCH_HEAD -- tests/test_dice_interface/test_dice_genetic.py

# --- BEFORE: production code from origin/main ---
git checkout origin/main -- dice_ml/explainer_interfaces/dice_genetic.py
python -m pytest tests/test_dice_interface/test_dice_genetic.py::TestMateContinuousMutation -q || echo \"BEFORE: AssertionError 'pinned to the lower bound' (expected)\"
# Expected: 1 failed — 'samples > 50.0' is False because every mutation collapsed to feature_range[name][0]

# --- AFTER: fix applied ---
git checkout FETCH_HEAD -- dice_ml/explainer_interfaces/dice_genetic.py
python -m pytest tests/test_dice_interface/test_dice_genetic.py::TestMateContinuousMutation -q
# Expected: 1 passed — mutations now spread across the declared [0, 100] range

What I ran locally

• New test passes on this branch
• New test fails on origin/main with all 200 samples == 0.0 (every mutation pinned to the lower bound of the feature range, exactly as the typo dictates)
• Direct line-by-line diff with adjacent np.random.uniform(low, high) call sites confirms the fix matches the established pattern in this file

Edge cases

call site	feature_range	mutation distribution
BEFORE: continuous mutation in mate	[0, 100]	all samples collapsed to 0.0
AFTER: continuous mutation in mate	[0, 100]	uniform over [0, 100]
Categorical mutation in mate	unchanged — uses np.random.choice	unchanged
Continuous mutation in do_random_init	already correct (low, high)	unchanged
Continuous mutation in do_KD_init	already correct (low, high)	unchanged

AI disclosure

This change was prepared with the assistance of Claude (Anthropic).
The author reviewed every line and is responsible for the final result.