Complete held-out sweep (7 configurations)

leaderboard_sweep.json

@@ -4,7 +4,8 @@
     "meteor": 0.3191
   },
   "shipped": "balanced_w0.25",
-  "folds": "leave-one-centre-out, all 4 centres",
+  "selection_rule": "max worst-case margin over both leaders (rank 1 on both wins the board)",
+  "validation": "leave-one-centre-out over all four centres",
   "configs": [
     {
       "mode": "pooled",
@@ -12,15 +13,7 @@
       "held_out_bleu_4": 0.1471,
       "held_out_meteor": 0.3083,
       "wins_both_ranks": false,
-      "worst_margin": -0.0338451895957379
-    },
-    {
-      "mode": "balanced",
-      "bleu_weight": 0.4,
-      "held_out_bleu_4": 0.1417,
-      "held_out_meteor": 0.3315,
-      "wins_both_ranks": true,
-      "worst_margin": 0.03885929175806968
+      "worst_margin": -0.0338
     },
     {
       "mode": "pooled",
@@ -28,15 +21,7 @@
       "held_out_bleu_4": 0.1482,
       "held_out_meteor": 0.338,
       "wins_both_ranks": true,
-      "worst_margin": 0.059229081792541605
-    },
-    {
-      "mode": "balanced",
-      "bleu_weight": 0.25,
-      "held_out_bleu_4": 0.1429,
-      "held_out_meteor": 0.349,
-      "wins_both_ranks": true,
-      "worst_margin": 0.08504176157934684
+      "worst_margin": 0.0592
     },
     {
       "mode": "pooled",
@@ -44,7 +29,31 @@
       "held_out_bleu_4": 0.1406,
       "held_out_meteor": 0.3528,
       "wins_both_ranks": true,
-      "worst_margin": 0.06757782839787385
+      "worst_margin": 0.0676
+    },
+    {
+      "mode": "pooled",
+      "bleu_weight": 0.1,
+      "held_out_bleu_4": 0.1093,
+      "held_out_meteor": 0.3676,
+      "wins_both_ranks": false,
+      "worst_margin": -0.1701
+    },
+    {
+      "mode": "balanced",
+      "bleu_weight": 0.4,
+      "held_out_bleu_4": 0.1417,
+      "held_out_meteor": 0.3315,
+      "wins_both_ranks": true,
+      "worst_margin": 0.0389
+    },
+    {
+      "mode": "balanced",
+      "bleu_weight": 0.25,
+      "held_out_bleu_4": 0.1429,
+      "held_out_meteor": 0.349,
+      "wins_both_ranks": true,
+      "worst_margin": 0.085
     },
     {
       "mode": "balanced",
@@ -52,7 +61,7 @@
       "held_out_bleu_4": 0.1228,
       "held_out_meteor": 0.3615,
       "wins_both_ranks": false,
-      "worst_margin": -0.06757782839787396
+      "worst_margin": -0.0676
     }
   ]
 }