v4.23.0b14

Author: songololo

analysis/sampling/paper/main.tex

@@ -14,7 +14,7 @@
 1\,km & \multicolumn{2}{c}{exact} & \multicolumn{2}{c}{exact} & \multicolumn{2}{c}{exact} & \multicolumn{2}{c}{exact} & \multicolumn{2}{c}{exact} \\
 2\,km & \multicolumn{2}{c}{exact} & \multicolumn{2}{c}{exact} & \multicolumn{2}{c}{exact} & \multicolumn{2}{c}{exact} & \multicolumn{2}{c}{exact} \\
 5\,km & \multicolumn{2}{c}{exact} & \multicolumn{2}{c}{exact} & 1{,}506 & 58.7\% & 847 & 33.0\% & 542 & 21.1\% \\
-10\,km & \multicolumn{2}{c}{exact} & 3{,}822 & 37.3\% & 1{,}698 & 16.6\% & 955 & 9.3\% & 611 & 6.0\% \\
+10\,km & \multicolumn{2}{c}{exact} & 3{,}822 & 37.3\% & 1{,}699 & 16.6\% & 955 & 9.3\% & 611 & 6.0\% \\
 20\,km & 17{,}022 & 41.5\% & 4{,}255 & 10.4\% & 1{,}891 & 4.6\% & 1{,}063 & 2.6\% & 680 & 1.7\% \\
 \bottomrule
 \end{tabular}

analysis/sampling/paper/tables/tab1_ew_comparison.tex

@@ -19,7 +19,7 @@
 
 \vspace{0.5em}
 \footnotesize
-Network: Greater London Area. Deterministic distance-based schedule:
+Network: Greater London Authority, \glaNnodes{} nodes. Deterministic distance-based schedule:
 same $p$ for both metrics at each distance.
 Subscripts: $c$ = closeness, $b$ = betweenness.
 \end{table}

analysis/sampling/paper/tables/tab2_validation.tex

@@ -19,6 +19,7 @@
 
 \vspace{0.5em}
 \footnotesize
-Network: Greater Madrid, 64,609 nodes. Deterministic distance-based schedule:
+Network: Greater Madrid, \madridNnodes{} nodes. Deterministic distance-based schedule:
 same $p$ for both metrics at each distance.
+Subscripts: $c$ = closeness, $b$ = betweenness.
 \end{table}

analysis/sampling/paper/tables/tab4_madrid_validation.tex

@@ -1,6 +1,6 @@
 \begin{table}[htbp]
 \centering
-\caption{Deterministic sampling schedule by analysis distance ($\varepsilon = 0.06$, $\delta = 0.1$, canonical grid spacing $s = 175\,\text{m}$). Both closeness and betweenness use the same schedule. Sampling is exact ($p = 100\%$) for distances up to 4\,km. Speedup shown is theoretical ($1/p$); measured wall-clock speedup is lower at near-unity $p$ due to sampling overhead and emerges at 10\,km and beyond.}
+\caption{Deterministic sampling schedule by analysis distance ($\varepsilon = 0.06$, $\delta = 0.1$, canonical grid spacing $s = 175\,\text{m}$). Both closeness and betweenness use the same schedule. Sampling is exact ($p = 100\%$) for distances up to 2\,km. Speedup shown is theoretical ($1/p$); measured wall-clock speedup is lower at near-unity $p$ due to sampling overhead and emerges at 10\,km and beyond.}
 \label{tab:distance_lookup}
 \begin{tabular}{rrrrr}
 \toprule

analysis/sampling/paper/tables/tab_distance_lookup.tex

@@ -305,7 +305,7 @@ def fmt_pct(p):
         d_str = f"{int(row['distance'] // 1000)}\\,km" if row['distance'] >= 1000 else f"{row['distance']}\\,m"
         latex += d_str
         for eps in EPSILON_TARGETS:
-            k = int(row[f"k_{eps}"])
+            k = math.ceil(row[f"k_{eps}"])
             p = row[f"p_{eps}"]
             if p >= 1.0:
                 latex += " & \\multicolumn{2}{c}{exact}"

analysis/sampling/scripts/01_analyse_synthetic.py

@@ -562,7 +562,7 @@ def generate_validation_table(summary_df: pd.DataFrame):
 
 \vspace{0.5em}
 \footnotesize
-Network: Greater London Area. Deterministic distance-based schedule:
+Network: Greater London Authority, \glaNnodes{} nodes. Deterministic distance-based schedule:
 same $p$ for both metrics at each distance.
 Subscripts: $c$ = closeness, $b$ = betweenness.
 \end{table}

analysis/sampling/scripts/02_validate_gla.py

@@ -559,15 +559,15 @@ def generate_validation_table(df: pd.DataFrame, n_nodes: int | None):
         latex += f"{p_pct} & {rho_c} & {spd_c} & "
         latex += f"& {rho_b} & {spd_b} \\\\\n"
 
-    n_nodes_str = f"{n_nodes:,}" if n_nodes else r"\texttildelow 99,000"
-    latex += rf"""\bottomrule
-\end{{tabular}}
+    latex += r"""\bottomrule
+\end{tabular}
 
-\vspace{{0.5em}}
+\vspace{0.5em}
 \footnotesize
-Network: Greater Madrid, {n_nodes_str} nodes. Deterministic distance-based schedule:
+Network: Greater Madrid, \madridNnodes{} nodes. Deterministic distance-based schedule:
 same $p$ for both metrics at each distance.
-\end{{table}}
+Subscripts: $c$ = closeness, $b$ = betweenness.
+\end{table}
 """
 
     output_path = TABLES_DIR / "tab4_madrid_validation.tex"

analysis/sampling/scripts/03_validate_madrid.py

@@ -102,7 +102,7 @@ def generate_macros() -> str:
     for dist in [1000, 2000, 5000, 10000, 20000]:
         p = compute_distance_p(dist, epsilon=PAPER_EPSILON_CLOSENESS, delta=HOEFFDING_DELTA)
         r = math.pi * dist**2 / GRID_SPACING**2
-        k = math.log(2 * r / HOEFFDING_DELTA) / (2 * PAPER_EPSILON_CLOSENESS**2)
+        k = math.ceil(math.log(2 * r / HOEFFDING_DELTA) / (2 * PAPER_EPSILON_CLOSENESS**2))
         speedup = 1.0 / p if p < 1.0 else 1.0
         distance_scenarios[dist] = {"p": p, "k": k, "canonical_reach": r, "speedup": speedup}
 
@@ -349,7 +349,7 @@ def generate_tab_distance_lookup() -> str:
         + f"{GRID_SPACING:.0f}"
         + r"\,\text{m}$). "
         r"Both closeness and betweenness use the same schedule. "
-        r"Sampling is exact ($p = 100\%$) for distances up to 4\,km. "
+        r"Sampling is exact ($p = 100\%$) for distances up to 2\,km. "
         r"Speedup shown is theoretical ($1/p$); measured wall-clock speedup is lower "
         r"at near-unity $p$ due to sampling overhead and emerges at 10\,km and beyond.}"
     )
@@ -359,10 +359,10 @@ def generate_tab_distance_lookup() -> str:
     lines.append(r"\textbf{Distance} & \textbf{Canonical reach} & \textbf{$k$} & \textbf{$p$ (\%)} & \textbf{Speedup ($1/p$)} \\")
     lines.append(r"\midrule")
     for d, r, p, speedup in rows:
-        k = math.log(2 * r / delta) / (2 * eps**2) if r > 0 else 0
+        k = math.ceil(math.log(2 * r / delta) / (2 * eps**2)) if r > 0 else 0
         d_str = f"{d // 1000}\\,km"
         r_str = f"{r:,.0f}".replace(",", "{,}")
-        k_str = f"{k:,.0f}".replace(",", "{,}")
+        k_str = f"{k:,}".replace(",", "{,}")
         p_str = f"{p * 100:.0f}" if p >= 1.0 else f"{p * 100:.1f}"
         spd_str = "1$\\times$" if p >= 1.0 else f"{speedup:.1f}$\\times$"
         lines.append(f"{d_str} & {r_str} & {k_str} & {p_str} & {spd_str} \\\\")