[Feat] Add adaptive graph construction (k-NN + Delaunay)

src/weather_model_graphs/graph_adaptive.py

@@ -0,0 +1,64 @@
+"""
+Adaptive graph construction utilities for non-uniform input grids.
+
+Supports:
+- k-NN graph construction (for sparse data)
+- Delaunay triangulation (for irregular grids)
+- Degree diagnostics for graph analysis
+"""
+
+from typing import List, Tuple
+import numpy as np
+from sklearn.neighbors import NearestNeighbors
+from scipy.spatial import Delaunay
+
+
+def build_knn_graph(coords: np.ndarray, k: int = 5) -> List[Tuple[int, int]]:
+    if coords.shape[0] < k:
+        raise ValueError("Number of points must be >= k")
+
+    nbrs = NearestNeighbors(n_neighbors=k)
+    nbrs.fit(coords)
+    _, indices = nbrs.kneighbors(coords)
+
+    edges = []
+    for i, neighbors in enumerate(indices):
+        for j in neighbors:
+            if i != j:
+                edges.append((i, j))
+
+    return edges
+
+
+def build_delaunay_graph(coords: np.ndarray) -> List[Tuple[int, int]]:
+    if coords.shape[0] < 3:
+        raise ValueError("Need at least 3 points")
+
+    tri = Delaunay(coords)
+    edges = set()
+
+    for simplex in tri.simplices:
+        for i in range(3):
+            u = simplex[i]
+            v = simplex[(i + 1) % 3]
+            edges.add((u, v))
+            edges.add((v, u))
+
+    return list(edges)
+
+
+def compute_degree(edges: List[Tuple[int, int]], num_nodes: int) -> List[int]:
+    degree = [0] * num_nodes
+
+    for u, _ in edges:
+        degree[u] += 1
+
+    return degree
+
+
+def degree_statistics(degree: List[int]) -> dict:
+    return {
+        "min": min(degree),
+        "max": max(degree),
+        "mean": sum(degree) / len(degree),
+    }