用高维投影实现无需训练的快速分类，性能媲美传统方法。
VoodooNet: Achieving Analytic Ground States via High-Dimensional Random Projections
- 通过高维空间投影和伪逆直接求解输出层，跳过迭代优化。
- MNIST准确率达98.10%，Fashion-MNIST达86.63%，远超10轮SGD基线。
- 适合对延迟敏感的边缘AI场景，可实现即时模型推理。
我们提出VoodooNet，一种非迭代神经架构，将随机梯度下降（SGD）范式替换为通过银河扩张实现的闭式解析解。通过将输入流形投影至高维、高熵的“银河”空间（维度d ≫ 784），我们证明复杂特征可在无反向传播热力学代价下被解耦。利用Moore-Penrose伪逆在单步内求解输出层，VoodooNet在MNIST上达到98.10%的分类准确率，在Fashion-MNIST上达到86.63%。值得注意的是，其在Fashion-MNIST上的表现超越10轮SGD基线（84.41%），同时训练时间减少数个数量级。我们观察到准确率与维度间近乎对数的缩放关系，表明性能取决于“银河”体积而非迭代优化。该“魔法帽”方法为实时边缘AI开辟了新路径，使传统训练阶段被绕过，实现即时流形发现。
原文摘要 · Abstract (English)
We present VoodooNet, a non-iterative neural architecture that replaces the stochastic gradient descent (SGD) paradigm with a closed-form analytic solution via Galactic Expansion. By projecting input manifolds into a high-dimensional, high-entropy "Galactic" space ($d \gg 784$), we demonstrate that complex features can be untangled without the thermodynamic cost of backpropagation. Utilizing the Moore-Penrose pseudoinverse to solve for the output layer in a single step, VoodooNet achieves a classification accuracy of \textbf{98.10\% on MNIST} and \textbf{86.63\% on Fashion-MNIST}. Notably, our results on Fashion-MNIST surpass a 10-epoch SGD baseline (84.41\%) while reducing the training time by orders of magnitude. We observe a near-logarithmic scaling law between dimensionality and accuracy, suggesting that performance is a function of "Galactic" volume rather than iterative refinement. This "Magic Hat" approach offers a new frontier for real-time Edge AI, where the traditional training phase is bypassed in favor of instantaneous manifold discovery.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。