Datasets:
File size: 2,210 Bytes
91034b5 1a6e388 91034b5 1a6e388 91034b5 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 | """Attach one-hot node features from nodes_ML to the saved PyG graph.
Each feature group is one categorical column; a node has exactly one 1 per
group, so every row sums to the number of groups.
"""
import duckdb
import torch
from node2vec_model import load_graph
from paths import DB_PATH, GRAPH_PATH
DB = DB_PATH
GRAPH = GRAPH_PATH
# column expression -> group name. tissue is field 2 of the pipe-delimited
# factor string, e.g. "(Camellia sinensis|leaf||experimental sample|...)".
GROUPS = {
"n.Study": "Study",
"n.Analytics": "Analytics",
"n.species": "species",
"n.factor1_org": "factor1_org",
"n.factor2_org": "factor2_org",
"n.color_group": "color_group",
"n.color_group_upper1": "color_group_upper1",
"split_part(trim(n.factor1, '()'), '|', 2)": "tissue1",
"split_part(trim(n.factor2, '()'), '|', 2)": "tissue2",
}
data = load_graph(GRAPH)
con = duckdb.connect(DB, read_only=True)
select = ", ".join(f'{e} AS "{name}"' for e, name in GROUPS.items())
rows = con.execute(f"""
SELECT g.id, {select}
FROM graph_nodes g JOIN nodes_ML n ON n.id = g.id
ORDER BY g.idx
""").fetchall()
ids = [r[0] for r in rows]
assert ids == data.node_id, "node order does not match the saved graph"
blocks, names, bounds = [], [], {}
offset = 0
for col, group in enumerate(GROUPS.values(), start=1):
values = sorted({r[col] for r in rows})
index = {v: i for i, v in enumerate(values)}
block = torch.zeros(len(rows), len(values))
block[torch.arange(len(rows)), torch.tensor([index[r[col]] for r in rows])] = 1.0
blocks.append(block)
names += [f"{group}={v}" for v in values]
bounds[group] = (offset, offset + len(values))
offset += len(values)
print(f" {group:<20} {len(values):>4} dims cols {bounds[group]}")
x = torch.cat(blocks, dim=1)
data.x = x
data.feature_names = names
data.feature_groups = bounds
torch.save(data, GRAPH)
print(f"\nx = {tuple(x.shape)} {x.dtype}")
print(f"row sums: min={x.sum(1).min():.0f} max={x.sum(1).max():.0f} "
f"(expected {len(GROUPS)} = one 1 per group)")
print(f"all-zero columns: {(x.sum(0) == 0).sum().item()}")
print(f"density: {x.mean():.4f}")
print(f"saved -> {GRAPH}")
print(data)
|