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

节点 n57

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 60aec6b2bf: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 3ed9cbb827。改动的文件：solution/METHOD.md +2 −2、solution/run.py +132 −35

diff --git a/solution/METHOD.md b/solution/METHOD.mdindex 601de6d..7b826a5 100644--- a/solution/METHOD.md+++ b/solution/METHOD.md@@ -1,4 +1,4 @@ ## 改了什么-将 with-replace 加权抽样改为按类型分层的 without-replace 抽样：先按家族权重计算每型目标细胞数，再在每型内无放回抽样（不足时取全部），剩余名额从未选细胞中补齐。不加任何表达位移。理由：Program 1 用同样的分层无放回抽样在 de_recovery 上得到 51.33（vs 本节点 with-replace 的 49.35），且无放回避免了重复细胞导致的经验分布失真，应同时有利于 covariation。+在父节点（简单无放回加权抽样，无 binning，无位移）基础上做三项改动：(1) 加入距离分箱分层抽样（N_BINS=20, MIN_CELLS_FOR_BINNING=30），保持型内到质心距离分布形状，预期提升 covariation；(2) 权重从 1.6/0.15/0.9 调为 1.5/0.2/0.95（Program 1 最佳配置）；(3) 当有两个官方输入阶段时，加 EB 收缩伪批量位移（ALPHA=0.04），预期提升 direction。父节点 ANALYSIS 指出"位移破坏 covariation"的结论是在 with-replace 下得出的，无放回+分箱条件下未必成立，故尝试小幅位移。 ## 用到的知识与出处-Program 1 实验结果（de_recovery 51.33、direction 54.01，分层无放回 + EB 位移）；父节点 9 配置（权重 1.6/0.15/0.9）；第 0 轮反馈证明 EB 位移在 with-replace 下破坏 covariation，故本轮只用无放回抽样、不加位移。+Program 1（score 52.65，分箱+权重 1.5/0.2/0.95）；Inspiration 1（score 51.56，ALPHA=0.04 EB 位移）；父节点 30 ANALYSIS 建议"在无放回条件下重新尝试小幅 EB 位移"；方法卡 T1-02（收缩系数 α 为超参）。diff --git a/solution/run.py b/solution/run.pyindex 05ac101..3d0d028 100644--- a/solution/run.py+++ b/solution/run.py@@ -1,12 +1,10 @@ #!/usr/bin/env python3-"""Composition reweighting: weighted sampling from latest official input stage.+"""Stratified reweighted copy + EB-shrunk delta on two-stage views. -Cell types are classified into families by regex patterns on their names.-Heart-related types are upweighted, surface ectoderm / extraembryonic /-neural types are downweighted or dropped. Expression values are unchanged.--Uses include_external=False to avoid mismatched-label external stages in-proxy2. Falls back to all stages if no official inputs exist (X3).+Combines distance-bin stratified sampling (preserves within-type shape,+boosts covariation) with a small EB-shrunk pseudobulk delta when two+official stages are available (boosts direction). Weights 1.5/0.2/0.95+match the best-scoring configuration. Without-replace sampling throughout. """  from __future__ import annotations@@ -15,7 +13,9 @@ import argparse 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 +26,10 @@ from src.task1_temporal.view_io import (     write_prediction, ) +ALPHA = 0.04+N_BINS = 20+MIN_CELLS_FOR_BINNING = 30+ _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 +47,121 @@ _GUT_PAT = re.compile(  def family_weight(name: str) -> float:     if _HEART_PAT.search(name):-        return 1.6+        return 1.5     if _REDUCED_PAT.search(name):-        return 0.15+        return 0.2     if _GUT_PAT.search(name):-        return 0.9+        return 0.95     return 1.0  +def _submatrix(X, mask):+    sub = X[mask]+    if sp.issparse(sub):+        return np.asarray(sub.todense())+    return np.asarray(sub)+++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 = float(np.mean(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 stratified_sample(labels, X, n_target, rng, tw):+    unique, counts = np.unique(labels, return_counts=True)+    n_total = len(labels)+    if n_target >= n_total:+        return np.arange(n_total)+    wc = counts.astype(np.float64) * tw+    props = wc / wc.sum()+    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]+        if n_take >= len(type_idx):+            indices.append(type_idx)+            continue+        if len(type_idx) > MIN_CELLS_FOR_BINNING and n_take >= 5:+            sub = _submatrix(X, type_idx)+            centroid = sub.mean(axis=0)+            dists = np.linalg.norm(sub - centroid, axis=1)+            bin_edges = np.quantile(dists, np.linspace(0, 1, N_BINS + 1)[1:-1])+            bin_ids = np.searchsorted(bin_edges, dists)+            bin_props = np.bincount(bin_ids, minlength=N_BINS) / len(type_idx)+            per_bin = np.floor(bin_props * n_take).astype(int)+            rem = n_take - per_bin.sum()+            fr = (bin_props * n_take) - per_bin+            for i in np.argsort(-fr)[:rem]:+                per_bin[i] += 1+            per_bin = np.minimum(per_bin, np.bincount(bin_ids, minlength=N_BINS))+            chosen = []+            for b in range(N_BINS):+                b_idx = type_idx[bin_ids == b]+                nb = min(per_bin[b], len(b_idx))+                if nb > 0:+                    chosen.append(rng.choice(b_idx, size=nb, replace=False))+            if chosen:+                sel = np.concatenate(chosen)+            else:+                sel = rng.choice(type_idx, size=n_take, replace=False)+            if len(sel) < n_take:+                leftover = np.setdiff1d(type_idx, sel)+                extra = rng.choice(leftover, size=n_take - len(sel), replace=False)+                sel = np.concatenate([sel, extra])+            indices.append(sel)+        else:+            indices.append(rng.choice(type_idx, size=n_take, replace=False))+    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()     parser.add_argument("--data", required=True)@@ -70,32 +181,18 @@ def main() -> None:     rng = np.random.default_rng(args.seed)     n_target = target_n_cells(manifest, last.n_obs) -    unique_labels, counts = np.unique(labels, return_counts=True)-    n_total = len(labels)-    raw_w = np.array([family_weight(str(u)) for u in unique_labels], dtype=np.float64)-    target_props = (counts / n_total) * raw_w-    target_props /= target_props.sum()-    n_per_type = np.floor(target_props * n_target).astype(int)-    remainder = n_target - n_per_type.sum()-    frac = (target_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_labels, n_per_type):-        if n_take <= 0:-            continue-        type_idx = np.where(labels == u)[0]-        if n_take >= len(type_idx):-            indices.append(type_idx)-        else:-            indices.append(rng.choice(type_idx, size=n_take, replace=False))-    rows = np.concatenate(indices)-    if len(rows) < n_target:-        remaining = np.setdiff1d(np.arange(n_total), rows)-        extra = rng.choice(remaining, size=n_target - len(rows), replace=False)-        rows = np.concatenate([rows, extra])+    unique_labels = np.unique(labels)+    tw = np.array([family_weight(str(u)) for u in unique_labels], dtype=np.float64)+    rows = stratified_sample(labels, last.X, n_target, rng, tw)     X = last.X[rows]++    if len(stages) >= 2:+        prev = read_stage(args.data, stages[-2], genes)+        prev_labels = labels_of(prev)+        deltas = eb_shrunk_deltas(prev.X, prev_labels, 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

SEARCH
```python
#!/usr/bin/env python3
"""Composition reweighting: weighted sampling from latest official input stage.

Cell types are classified into families by regex patterns on their names.
Heart-related types are upweighted, surface ectoderm / extraembryonic /
neural types are downweighted or dropped. Expression values are unchanged.

Uses include_external=False to avoid mismatched-label external stages in
proxy2. Falls back to all stages if no official inputs exist (X3).
"""

from __future__ import annotations

import argparse
import re

import numpy as np
from scipy import sparse as sp

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

N_BINS = 20
MIN_CELLS_FOR_BINNING = 30

_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",
    re.IGNORECASE,
)
_REDUCED_PAT = re.compile(