Show the value line plots and tree graphs in the card
Browse files
README.md
CHANGED
|
@@ -115,14 +115,35 @@ by ranking actions.
|
|
| 115 |
|
| 116 |
## Figures
|
| 117 |
|
| 118 |
-
Two seeds per task
|
| 119 |
-
|
| 120 |
-
|
| 121 |
-
|
| 122 |
-
|
| 123 |
-
|
| 124 |
-
|
| 125 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 126 |
|
| 127 |
## Reading it
|
| 128 |
|
|
|
|
| 115 |
|
| 116 |
## Figures
|
| 117 |
|
| 118 |
+
Two seeds per task — 26 of each, as SVG under `figures/` and as PNG under `figures/png/`.
|
| 119 |
+
|
| 120 |
+
### The root's value, as the evidence arrives
|
| 121 |
+
|
| 122 |
+
The running estimate of `V(root)` after each of the 48 continuations, with a ±1 s.e. band.
|
| 123 |
+
The band assumes independent draws; a tree's are not, since simulations share prefixes — so
|
| 124 |
+
it is honest about when an estimate has settled and optimistic as a confidence interval.
|
| 125 |
+
|
| 126 |
+
A task the policy is good at, and one it is not:
|
| 127 |
+
|
| 128 |
+

|
| 129 |
+
|
| 130 |
+

|
| 131 |
+
|
| 132 |
+
The estimate is noisy for the first handful of continuations and settles by roughly twenty —
|
| 133 |
+
which is what the budget of 48 was chosen against.
|
| 134 |
+
|
| 135 |
+
### The search tree, valued
|
| 136 |
+
|
| 137 |
+
Every node filled by its subtree value on a red→yellow→green ramp, sized by the endings
|
| 138 |
+
behind that value, and ringed if it ended. The ring is the tier; the fill is the value.
|
| 139 |
+
|
| 140 |
+

|
| 141 |
+
|
| 142 |
+
The shape is the point: progressive widening spends the budget near the root, so the tree is
|
| 143 |
+
wide and shallow — 8 children at depth 1, 40 at depth 2 — and each of those runs out to the
|
| 144 |
+
horizon as a spine of nodes that share one value until the ending that set it.
|
| 145 |
+
|
| 146 |
+

|
| 147 |
|
| 148 |
## Reading it
|
| 149 |
|