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

节点 n27

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 fcefb7bee5: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 d36a9858b2。改动的文件：solution/METHOD.md +2 −2、solution/run.py +61 −7

diff --git a/solution/METHOD.md b/solution/METHOD.mdindex 9641820..c4300a9 100644--- a/solution/METHOD.md+++ b/solution/METHOD.md@@ -1,4 +1,4 @@ ## 改了什么-回退第0轮的分层抽样和均值拉扯（导致covariation从51.96暴跌到44.55），恢复父节点的纯加权有放回抽样（不修改表达值，完美保留基因间协变结构）。在此基础上调整家族权重：heart 1.6→1.8、reduced 0.15→0.08、gut 0.9→0.85，更激进地上调心脏族、下调非心脏族，预期进一步提升cell_state而不损害covariation。+回退权重到已验证的1.6/0.15/0.9（1.8/0.08/0.85已证明在噪声平台上无收益），保留加权有放回抽样（covariation安全基线），新增经验贝叶斯收缩伪批量差值（eb_shrunk_deltas, ALPHA=0.03）：当有两个官方输入阶段时对抽样后的细胞施加按类型收缩的位移，提升direction和de_recovery；单阶段时退化为纯抽样。改用手动过滤official阶段（与Program 1一致），确保X3视图兼容。 ## 用到的知识与出处-第0轮实验反馈（均值拉扯beta=0.05导致covariation -7.4，证实任何表达值修改都破坏协变结构）；方向库T1-01（组成重加权，种子heart_jcf_peri用×1.6和×0.25）；方法卡T1标签节（E8.5独有名应降权）；父节点ANALYSIS建议扫参reduced权重（0.0/0.15/0.25）。+方法卡T1平移幅度节（收缩系数α为超参，不能默认α=1）；Program 1实验结果（eb_shrunk_deltas使direction +0.86、de_recovery +1.92）；父节点ANALYSIS（权重微调已枯竭，应转向结构性改变）；方向库T1-02（按类型伪批量位移+收缩系数）。diff --git a/solution/run.py b/solution/run.pyindex 115cde7..225b45a 100644--- a/solution/run.py+++ b/solution/run.py@@ -16,6 +16,9 @@ import re  import numpy as np +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,@@ -26,6 +29,8 @@ from src.task1_temporal.view_io import (     write_prediction, ) +ALPHA = 0.03+ _HEART_PAT = re.compile(     r"cm\b|cardi|heart|shf|endocard|oft\b|avc|ift\b|sv[- ]cm|v[- ]cm|"     r"proepicard|pericard|epicard|blood",@@ -43,14 +48,55 @@ _GUT_PAT = re.compile(  def family_weight(name: str) -> float:     if _HEART_PAT.search(name):-        return 1.8+        return 1.6     if _REDUCED_PAT.search(name):-        return 0.08+        return 0.15     if _GUT_PAT.search(name):-        return 0.85+        return 0.9     return 1.0  +def eb_shrunk_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+        n_last = int(last_mask.sum())+        if n_last == 0:+            continue+        last_sub = last_X[last_mask]+        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:+            arr = np.asarray(last_sub)+            mean_last = arr.mean(axis=0).ravel()+            var_last = arr.var(axis=0).ravel()+        var_last = np.maximum(var_last, 0.0)+        if t in prev_types:+            prev_mask = prev_labels == t+            n_prev = int(prev_mask.sum())+            prev_sub = prev_X[prev_mask]+            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:+                arr_p = np.asarray(prev_sub)+                mean_prev = arr_p.mean(axis=0).ravel()+                var_prev = arr_p.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)+            n_eff = min(n_prev, n_last)+            size_factor = n_eff / (n_eff + 50.0)+            shrink = delta**2 / (delta**2 + se2 + 1e-10)+            deltas[t] = alpha * size_factor * shrink * delta+        else:+            deltas[t] = np.zeros(last_X.shape[1])+    return deltas++ def main() -> None:     parser = argparse.ArgumentParser()     parser.add_argument("--data", required=True)@@ -61,11 +107,12 @@ def main() -> None:     manifest = load_manifest(args.data)     genes = panel_genes(args.data, manifest) -    stages = inputs_by_time(manifest, include_external=False)-    if not stages:-        stages = inputs_by_time(manifest)+    stages = inputs_by_time(manifest)+    official = [s for s in stages if s.get("source") != "external"]+    if not official:+        official = stages -    last = read_stage(args.data, stages[-1], genes)+    last = read_stage(args.data, official[-1], genes)     labels = labels_of(last)     rng = np.random.default_rng(args.seed)     n_target = target_n_cells(manifest, last.n_obs)@@ -77,6 +124,13 @@ def main() -> None:      rows = rng.choice(last.n_obs, size=n_target, replace=True, p=weights)     X = last.X[rows]++    if len(official) >= 2:+        prev = read_stage(args.data, official[-2], genes)+        deltas = eb_shrunk_deltas(prev.X, labels_of(prev), last.X, labels, ALPHA)+        del prev+        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 round 1 results, the E8.5-only type dropping was catastrophic (score 47.33 vs 51.12). The problem: in the proxy view, E8.5 IS the input stage, so dropping those types removes most cells. The method card's "don't output E8.5-only names" applies to the E10.5 final target, not the E9.5 proxy target.

Strategy: revert to Program 1 exactly (best at 52.15) — remove E85_ONLY, restore eb_shrunk_deltas with ALPHA=0.03.

<<<<<<< SEARCH
from src.task1_temporal.view_io import (
    inputs_by_time,
    labels_of,
    load_manifest,
    panel_genes,
    read_stage,
    target_n_cells,
    write_prediction,
)



_HEART_PAT = re.compile(
=======
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,
    target_n_cells,
    write_prediction,
)

ALPHA = 0.03

_HEART_PAT = re.compile(
>>>>>>> REPLACE

<<<<<<< SEARCH
_E85_ONLY = re.compile(
    r"^(LV[- ]CM|RV[- ]CM|Endothelium|Neural Tube|JCF|EXEM|Paraxial Mesoderm)$",