logbook.json (verbatim titles)

logbook.json

@@ -0,0 +1,70 @@
+{
+ "schema_version": 1,
+ "title": "Reproduction: Improved Stochastic Optimization of LogSumExp",
+ "emoji": "🧮",
+ "space_id": "snaykey/repro-logsumexp-opt",
+ "paper": {
+  "arxiv_id": "2509.24894",
+  "openreview_id": "TzQElzflxR"
+ },
+ "tags": [
+  "icml2026-repro",
+  "paper-TzQElzflxR"
+ ],
+ "updated_at": "2026-07-28T00:00:00+00:00",
+ "root": {
+  "slug": "index",
+  "title": "Reproduction: Improved Stochastic Optimization of LogSumExp",
+  "file": "pages/index.md",
+  "children": [
+   {
+    "slug": "executive-summary",
+    "title": "Executive summary",
+    "file": "pages/executive-summary/page.md",
+    "children": []
+   },
+   {
+    "slug": "claim-1-safe-kl-approximation-bound",
+    "title": "The paper defines a Safe KL divergence with a bounded-density constraint (Definition 2.1, Equations 3-4) whose induced approximation to LogSumExp satisfies F_ρ − O(ρ) ≤ F ≤ F_ρ (Proposition 2.4, Section 2).",
+    "file": "pages/claim-1-safe-kl-approximation-bound/page.md",
+    "children": []
+   },
+   {
+    "slug": "claim-2-softplus-conjugate-strong-convex-smooth",
+    "title": "The rescaled SoftPlus-based conjugate of the Safe KL function is proven to be ρ-strongly convex, with its own conjugate being (1/ρ)-smooth (Lemma 2.7, Section 2.2).",
+    "file": "pages/claim-2-softplus-conjugate-strong-convex-smooth/page.md",
+    "children": []
+   },
+   {
+    "slug": "claim-3-cvar-connection-limiting-cases",
+    "title": "The approximation family is shown to bound and connect to the CVaR functional, recovering CVaR and LogSumExp as limiting cases (Proposition 2.6, Section 2.1).",
+    "file": "pages/claim-3-cvar-connection-limiting-cases/page.md",
+    "children": []
+   },
+   {
+    "slug": "claim-4-eot-semidual-overflow-convergence",
+    "title": "In continuous entropy-regularized optimal transport experiments with regularization ε=0.01, the proposed semi-dual formulation (Equations 14-15) avoids the numerical overflow issues of the baseline and reaches a converged objective in roughly 10^4 iterations (Figure 2, Section 3.1).",
+    "file": "pages/claim-4-eot-semidual-overflow-convergence/page.md",
+    "children": []
+   },
+   {
+    "slug": "claim-5-kl-dro-california-housing-stepsize",
+    "title": "In KL-regularized distributionally robust optimization on California Housing, the proposed method converges with a stepsize of η=10⁻⁴ across batch sizes of 10, 100, and 1000, whereas the baseline requires a much smaller η=10⁻⁶ at batch size 10 (Figure 3, Section 3.2).",
+    "file": "pages/claim-5-kl-dro-california-housing-stepsize/page.md",
+    "children": []
+   },
+   {
+    "slug": "claim-6-uot-dro-mnist-stepsize",
+    "title": "In unbalanced optimal-transport-based DRO on MNIST with label noise, the proposed approach (Equation 22) converges faster than the baseline (Equation 21) using η=10⁻⁴ versus the baseline's required η=10⁻⁵ (Figure 4, Section 3.3).",
+    "file": "pages/claim-6-uot-dro-mnist-stepsize/page.md",
+    "children": []
+   },
+   {
+    "slug": "conclusion",
+    "title": "Conclusion",
+    "file": "pages/conclusion/page.md",
+    "children": []
+   }
+  ]
+ }
+}