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

节点 n5

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 303f5ba5d1: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 +82 −16

diff --git a/solution/METHOD.md b/solution/METHOD.mdnew file mode 100644index 0000000..fd78a00--- /dev/null+++ b/solution/METHOD.md@@ -0,0 +1,8 @@+## 改了什么+1. （继承第0轮）过滤外部输入阶段，修复 proxy2 崩溃；分层抽样保留类型比例；全局收缩 α=0.5。+2. 将均匀 α 收缩替换为逐基因经验贝叶斯收缩：对每个基因计算 delta 的标准误（基于两阶段各类型的方差和样本量），收缩因子 = delta² / (delta² + SE²)。信号强的基因获得接近 α 的位移，噪声基因几乎不动。这保留了基因间共变结构（covariation），因为不再对噪声基因施加随机方向的位移；同时改善 de_recovery，因为差异表达基因（信号强）获得更准确的位移。+3. 移除对 `type_deltas` 的依赖，改用自定义 `shrunk_type_deltas` 函数，内部处理稀疏矩阵。+## 用到的知识与出处+- 方法卡 k018：收缩系数 α∈[0,1]，逐基因经验贝叶斯收缩（"Optionally shrink per gene (empirical Bayes on the delta's variance) so noisy genes move less"）。+- James-Stein 收缩估计的一般原理：观测方差大的估计应向零收缩。+- CONTRACT 中 proxy2 外部输入阶段说明（第0轮已修复）。diff --git a/solution/run.py b/solution/run.pyindex f3a0f25..6a19df1 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@@ -16,18 +12,83 @@ import argparse  import numpy as np -from src.task1_temporal.baselines import shift_rows, type_deltas+from scipy import sparse as sp++from src.task1_temporal.baselines import shift_rows from src.task1_temporal.view_io import (     inputs_by_time,     labels_of,     load_manifest,     panel_genes,     read_stage,-    sample_rows,     target_n_cells,     write_prediction, ) +ALPHA = 0.5+++def shrunk_type_deltas(prev_X, prev_labels, last_X, last_labels, alpha):+    prev_types = set(np.unique(prev_labels))+    last_types = np.unique(last_labels)+    deltas = {}+    for t in last_types:+        last_mask = last_labels == t+        last_sub = last_X[last_mask]+        n_last = last_sub.shape[0]+        if sp.issparse(last_sub):+            mean_last = np.asarray(last_sub.mean(axis=0)).ravel()+            var_last = np.asarray(last_sub.multiply(last_sub).mean(axis=0)).ravel() - mean_last**2+        else:+            mean_last = np.asarray(last_sub).mean(axis=0).ravel()+            var_last = np.asarray(last_sub).var(axis=0).ravel()+        var_last = np.maximum(var_last, 0.0)+        if t in prev_types:+            prev_mask = prev_labels == t+            prev_sub = prev_X[prev_mask]+            n_prev = prev_sub.shape[0]+            if sp.issparse(prev_sub):+                mean_prev = np.asarray(prev_sub.mean(axis=0)).ravel()+                var_prev = np.asarray(prev_sub.multiply(prev_sub).mean(axis=0)).ravel() - mean_prev**2+            else:+                mean_prev = np.asarray(prev_sub).mean(axis=0).ravel()+                var_prev = np.asarray(prev_sub).var(axis=0).ravel()+            var_prev = np.maximum(var_prev, 0.0)+            delta = mean_last - mean_prev+            se2 = var_prev / max(n_prev, 1) + var_last / max(n_last, 1)+            shrink = delta**2 / (delta**2 + se2 + 1e-10)+            deltas[t] = alpha * shrink * delta+        else:+            deltas[t] = np.zeros(last_X.shape[1])+    return deltas+++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 +100,20 @@ 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 = shrunk_type_deltas(prev.X, labels_of(prev), last.X, labels, ALPHA)         del prev-        X = shift_rows(X, labels_of(last)[rows], deltas)+        X = shift_rows(X, labels[rows], deltas)     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

stream

Looking at the scores, direction (49.29) and covariation (42.94) are the weakest. The EB shrinkage helped covariation (+1.75), but I can further improve by:
1. Adding a minimum cell threshold to avoid noisy delta estimates for rare types
2. Masking out genes with near-zero expression (shifting them adds noise without signal)
3. Increasing alpha to 0.6 since the EB + mask provides sufficient regularization

<<<<<<< SEARCH
ALPHA = 0.5


def shrunk_type_deltas(prev_X, prev_labels, last_X, last_labels, alpha):
    prev_types = set(np.unique(prev_labels))
    last_types = np.unique(last_labels)
    deltas = {}
    for t in last_types:
        last_mask = last_labels == t
        last_sub = last_X[last_mask]
        n_last = last_sub.shape[0]
        if sp.issparse(last_sub):
            mean_last = np.asarray(last_sub.mean(axis=0)).ravel()
            var_last = np.asarray(last_sub.multiply(last_sub).mean(axis=0)).ravel() - mean_last**2
        else:
            mean_last = np.asarray(last_sub).mean(axis=0).ravel()
            var_last = np.asarray(last_sub).var(axis=0).ravel()
        var_last = np.maximum(var_last, 0.0)
        if t in prev_types:
            prev_mask = prev_labels == t
            prev_sub = prev_X[prev_mask]
            n_prev = prev_sub.shape[0]
            if sp.issparse(prev_sub):
                mean_prev = np.asarray(prev_sub.mean(axis=0)).ravel()
                var_prev = np.asarray(prev_sub.multiply(prev_sub).mean(axis=0)).ravel() - mean_prev**2
            else:
                mean_prev = np.asarray(prev_sub).mean(axis=0).ravel()
                var_prev = np.asarray(prev_sub).var(axis=0).ravel()
            var_prev = np.maximum(var_prev, 0.0)
            delta = mean_last - mean_prev
            se2 = var_prev / max(n_prev, 1) + var_last / max(n_last, 1)
            shrink = delta**2 / (delta**2 + se2 + 1e-10)
            deltas[t] = alpha * shrink * delta
        else: