跳到正文
arXiv cs.AI· Zahra Khodagholi, Niloofar Yousefi·· 4 小时前AI 评分20

可表示但未学到:编码秩与交互预测下限

Representable but Unlearned: Encoding Rank and the Interaction-Prediction Floor

AI 导读

研究者提出一种无需标签或拟合预测器的方法,从编码器与对比设计计算可达对比空间,并给出确定性解码器的经验平方误差下限。在 siRNA 面板上,图神经网络的训练专用特征掩码将 165 个测量状态合并为 90 类,把 140 个交互对比的秩降至 72,下限为 0.009980,占记录对比能量的 14.63%。该方法可定位编码施加的限制并追溯到具体特征,但本身不能预测重拟合收益或解释剩余误差。

正文 · 原文

View PDF HTML (experimental)

Abstract:Input encodings can prevent a predictor from reproducing measured interactions jointly, even when each interaction is individually representable. We compute the attainable contrast space from the encoder and contrast design, without labels or predictor fitting. Projecting measured contrasts onto this space gives an empirical squared-error floor for any deterministic decoder using the encoding. On an siRNA panel, a graph neural network's training-only feature mask merges 165 measured states into 90 classes, reducing the rank of 140 interaction contrasts to 72. The floor is 0.009980, or 14.63% of recorded contrast energy, the mean squared error of predicting zero for every contrast. Its 14.6% share of fitted interaction error adds little: predicted interactions are nearly zero, and their MSE of 0.068335 slightly exceeds that of the same control. Antisense-marginal shares are larger: 36.49% of recorded energy and 38.1% of fitted error. Three restored chemistry columns suffice for full rank. Removing the full mask eliminates the floor, but refitting improves interaction MSE by only 0.000017; the restored features remain absent from training inputs. Two further panels clarify the diagnostic's scope. A released splicing predictor encodes injectively and retains the full design rank of 1,986, so its floor is zero. For an independent count-fingerprint encoder on a lipid panel, 800 pairwise-distinct encoded contrast rows have rank 740, exposing 60 restrictions missed by duplicate-row checks. The diagnostic identifies encoding-imposed limits and can trace them to specific features. It separates representability from learned accuracy, but does not by itself predict refitting gains or explain the remaining error.
Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
Cite as: arXiv:2609.36208 [cs.LG]
  (or arXiv:2609.36208v2 [cs.LG] for this version)
  https://doi.org/10.48550/arXiv.2609.36208

arXiv-issued DOI via DataCite

Submission history

From: Zahra Khodagholi [view email]
[v1] Mon, 28 Sep 2026 20:09:11 UTC (856 KB)
[v2] Mon, 16 Nov 2026 23:26:31 UTC (1,011 KB)

来源:arXiv cs.AI · arxiv.org