Generate the missing heading map for autodiff.md (#142)

autodiff.md was the only file in the whole five-edition estate with
sections but no heading map in its frontmatter — 248 of 249 carry one.
Found while auditing whether a new deterministic headingMapCorrect check
would false-gate production PRs (QuantEcon/action-translation#148); it
turned out to be a genuine gap rather than a false positive.

The heading map bridges English heading IDs to translated headings, and
sync uses it to locate existing translations section by section. Without
it, the next sync of this lecture has nothing to match against.

Generated with `translate headingmap`, which compares source and target
sections by position and makes no LLM calls — 5 top-level sections, 16
entries, all matched cleanly (5s/5t). A full-edition dry run reports the
other 25 files unchanged, so this is exactly one file.

The state file's tool-version stamp moves 0.13.1 -> 0.22.0 as a side
effect of the write; that field records the CLI that last wrote the
state, so the bump is accurate.

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Author: mmcky

.translate/state/autodiff.md.yml

@@ -3,4 +3,4 @@ synced-at: "2026-04-08"
 model: claude-sonnet-4-6
 mode: NEW
 section-count: 5
-tool-version: 0.13.1
+tool-version: 0.22.0

lectures/autodiff.md

@@ -9,6 +9,25 @@ kernelspec:
   display_name: Python 3 (ipykernel)
   language: python
   name: python3
+translation:
+  title: ماجراهایی با مشتق‌گیری خودکار
+  headings:
+    Overview: مرور کلی
+    What is automatic differentiation?: مشتق‌گیری خودکار چیست؟
+    What is automatic differentiation?::Autodiff is not finite differences: مشتق‌گیری خودکار، تفاضل محدود نیست
+    What is automatic differentiation?::Autodiff is not symbolic calculus: مشتق‌گیری خودکار، حساب نمادین نیست
+    What is automatic differentiation?::Autodiff: مشتق‌گیری خودکار
+    Some experiments: برخی آزمایش‌ها
+    Some experiments::A differentiable function: یک تابع قابل مشتق‌گیری
+    Some experiments::Absolute value function: تابع قدر مطلق
+    Some experiments::Differentiating through control flow: مشتق‌گیری از طریق جریان کنترل
+    Some experiments::Differentiating through a linear interpolation: مشتق‌گیری از طریق درون‌یابی خطی
+    Gradient Descent: گرادیان کاهشی
+    Gradient Descent::A function for gradient descent: یک تابع برای گرادیان کاهشی
+    Gradient Descent::Simulated data: داده‌های شبیه‌سازی‌شده
+    Gradient Descent::Minimizing squared loss by gradient descent: کمینه‌سازی تابع زیان مربعات با گرادیان کاهشی
+    Gradient Descent::Adding a squared term: افزودن یک جمله مربعی
+    Exercises: تمرین‌ها
 ---
 
 # ماجراهایی با مشتق‌گیری خودکار