总览 · ← 返回运行 20261002-034201-search-t1-abc-r1-C-native

节点 n3

改了什么

方法说明?节点程序自带的 METHOD.md：这个程序做了什么、为什么。

来自 programs.git 85e30575ef:solution/METHOD.md

调研员的计划

代码改动?这个节点的程序和父节点程序的逐行差别：绿色是新增，红色是删除。

对比：父节点版本 5d0b878932。改动的文件：solution/METHOD.md +8 −0、solution/run.py +45 −15

diff --git a/solution/METHOD.md b/solution/METHOD.mdnew file mode 100644index 0000000..0f9313e--- /dev/null+++ b/solution/METHOD.md@@ -0,0 +1,8 @@+## 改了什么+1. 将基础采样限定为官方阶段（过滤 source=="external"），修复 proxy2 中从 Qiu E9.0 心脏细胞采样导致全胚预测崩溃的问题（proxy2 从 27→预期~50）。+2. 用按细胞类型分层抽样替代均匀随机抽样，保持类型比例从而改善 covariation 分数。+3. 对伪批量差值乘收缩系数 α=0.5（方法卡 k018 指出 α=1 低于 copy_last，需要收缩）。+## 用到的知识与出处+- 方法卡 k018（Damped per-type shift: shrinkage alpha）+- 任务书 CONTRACT 中关于 proxy2 外部输入阶段的说明（inputs_by_time 默认返回外部阶段、类型名不匹配）+- 方向库 T1-01（组成重加权）的思路用于分层采样保持比例diff --git a/solution/run.py b/solution/run.pyindex f3a0f25..abfc55a 100644--- a/solution/run.py+++ b/solution/run.py@@ -1,13 +1,9 @@ #!/usr/bin/env python3-"""pseudobulk_shift: latest stage + per-cell-type pseudobulk delta of the last step.+"""Stratified copy with shrunk pseudobulk delta on official stages only. -The delta is mean(last|type) - mean(prev|type) over the two latest inputs,-computed on the full stages and added once to a subsample of the latest stage-(clipped at 0). Types missing from the earlier stage are copied unchanged.--With a single input stage (T1 proxy: E8.5 only) there is no step to take a-delta from, so this falls back to copy_last with the same sampling. The proxy-therefore cannot tell this seed from copy_last; that gap is expected.+Fixes proxy2 collapse (was sampling external heart-only stage as base),+uses stratified sampling to preserve type proportions (covariation),+and applies shrinkage alpha=0.5 to deltas per method card k018. """  from __future__ import annotations@@ -23,11 +19,39 @@ from src.task1_temporal.view_io import (     load_manifest,     panel_genes,     read_stage,-    sample_rows,     target_n_cells,     write_prediction, ) +ALPHA = 0.5+++def stratified_sample(labels: np.ndarray, n_target: int, rng: np.random.Generator) -> np.ndarray:+    unique, counts = np.unique(labels, return_counts=True)+    n_total = len(labels)+    if n_target >= n_total:+        return np.arange(n_total)+    props = counts / n_total+    n_per_type = np.floor(props * n_target).astype(int)+    remainder = n_target - n_per_type.sum()+    frac = (props * n_target) - n_per_type+    top_idx = np.argsort(-frac)[:remainder]+    n_per_type[top_idx] += 1+    n_per_type = np.minimum(n_per_type, counts)+    indices = []+    for u, n_take in zip(unique, n_per_type):+        if n_take <= 0:+            continue+        type_idx = np.where(labels == u)[0]+        chosen = rng.choice(type_idx, size=n_take, replace=False)+        indices.append(chosen)+    result = np.concatenate(indices)+    if len(result) < n_target:+        remaining_idx = np.setdiff1d(np.arange(n_total), result)+        extra = rng.choice(remaining_idx, size=n_target - len(result), replace=False)+        result = np.concatenate([result, extra])+    return result+  def main() -> None:     parser = argparse.ArgumentParser()@@ -39,15 +63,21 @@ def main() -> None:     manifest = load_manifest(args.data)     genes = panel_genes(args.data, manifest)     stages = inputs_by_time(manifest)-    last = read_stage(args.data, stages[-1], genes)+    official = [s for s in stages if s.get("source") != "external"]+    if not official:+        official = stages+    last = read_stage(args.data, official[-1], genes)+    labels = labels_of(last)     rng = np.random.default_rng(args.seed)-    rows = sample_rows(last.n_obs, target_n_cells(manifest, last.n_obs), rng)+    n_target = target_n_cells(manifest, last.n_obs)+    rows = stratified_sample(labels, n_target, rng)     X = last.X[rows]-    if len(stages) >= 2:-        prev = read_stage(args.data, stages[-2], genes)-        deltas = type_deltas(prev.X, labels_of(prev), last.X, labels_of(last))+    if len(official) >= 2:+        prev = read_stage(args.data, official[-2], genes)+        deltas = type_deltas(prev.X, labels_of(prev), last.X, labels)         del prev-        X = shift_rows(X, labels_of(last)[rows], deltas)+        shrunk = {k: v * ALPHA for k, v in deltas.items()}+        X = shift_rows(X, labels[rows], shrunk)     write_prediction(X, genes, args.out, seed=args.seed)  

调研来源?调研员查到并用到的知识条目和文献检索结果（只列标题和编号）。

没有记录调研来源。

分析结果?分析员写的 ANALYSIS.json：改了什么、各组分数怎么变、假设是否成立、经验和下一步建议。

对话摘要?每个角色和大模型对话的统计：轮数、工具调用、用时、token 数和最后的回答摘录；原始记录只给路径。

只给统计和最后回答的摘录；完整对话请到原始记录位置里列出的文件看。

分析员

ANALYSIS.json written

native

审查员

REVIEW.json written

stream

The crash is because `X` is a sparse matrix. I need to densify it before kNN smoothing. Also, cKDTree in 32k dimensions is impractical — I'll use highly variable genes for neighbor finding.

<<<<<<< SEARCH
def knn_smooth(X: np.ndarray, labels: np.ndarray, k: int = K_NEIGHBORS, blend: float = BLEND) -> np.ndarray:
    X_out = X.copy()
    for ct in np.unique(labels):
        mask = labels == ct
        n_ct = int(mask.sum())
        if n_ct <= k + 1:
            continue
        X_ct = np.asarray(X[mask], dtype=np.float64)
        tree = cKDTree(X_ct)
        kk = min(k + 1, n_ct)
        _, idx = tree.query(X_ct, k=kk)
        neighbor_mean = X_ct[idx[:, 1:]].mean(axis=1)
        X_out[mask] = (1.0 - blend) * X_ct + blend * neighbor_mean
    return X_out.astype(np.float32)
=======
def knn_smooth(X, labels: np.ndarray, k: int = K_NEIGHBORS, blend: float = BLEND) -> np.ndarray:
    if hasattr(X, "toarray"):
        X = X.toarray()
    X = np.asarray(X, dtype=np.float32)
    var = X.var(axis=0)
    top_genes = np.argsort(-var)[:500]
    X_out = X.copy()
    for ct in np.unique(labels):
        mask = labels == ct
        n_ct = int(mask.sum())
        if n_ct <= k + 1:
            continue
        sub = X[np.ix_(mask, top_genes)].astype(np.float64)
        tree = cKDTree(sub)
        kk = min(k + 1, n_ct)
        _, idx = tree.query(sub, k=kk)
        X_ct = X[mask].astype(np.float64)
        neighbor_mean = X_ct[idx[:, 1:]].mean(axis=1)
        X_out[mask] = ((1.0 - blend) * X_ct + blend * neighbor_mean).astype(np.float32)
    return X_out
>>>>>>> REPLACE