TinyCast用14万参数实现零样本概率预测，靠计算周期性而非学习。
TinyCast: Probabilistic Zero-Shot Forecasting with Computed Periodicity
- 不依赖注意力机制，通过谱检测提取主周期并折叠上下文
- 在概率精度上突破小模型边界，仅146,505参数即达顶尖性能
- 适合嵌入式设备部署，支持整数量化与端到端推理
我们提出TinyCast，一种无需注意力机制的零样本预测模型，仅用146,505个参数，在此规模下认为上下文的周期结构应被计算而非学习。一个零参数的谱检测器识别主导周期，上下文按相位折叠后，由扩张卷积编码器与块自回归分位数解码器建模剩余部分。其参数量小于所有可确认参数量的GIFT-Eval零样本模型。在概率准确性上定义了小模型的性能边界：在声明无测试数据泄露的零样本模型中，它是唯一低于140万参数且能输出预测分布的模型，所有表现更优的模型参数量至少与此相当。在Chronos-ZS和fev-bench上，所有领先神经模型的参数量均至少是它的28倍。由于前向路径仅含卷积与矩阵乘法，可导出为静态INT8格式，在无需信号级拟合的情况下于嵌入式设备端到端完成预测。
原文摘要 · Abstract (English)
We introduce TinyCast, an attention-free zero-shot forecaster that emits a predictive distribution from 146,505 parameters, on the premise that at this size the periodic structure of a context is worth computing rather than learning. A zero-parameter spectral detector supplies the dominant periods, the context is folded on their phase, and a dilated convolutional encoder and a block-autoregressive quantile decoder model the rest. It is smaller than every zero-shot entry on the GIFT-Eval board whose parameter count can be established. On probabilistic accuracy it defines the size-accuracy frontier. Among zero-shot entries declaring no test-data leakage it is the only one below 1.4M parameters that emits a predictive distribution, and every entry scoring better carries at least that budget. On Chronos-ZS and fev-bench every neural model ahead of it carries at least 28 times its parameters. Because the mixing path is convolutions and matrix multiplications only, it exports to static INT8 and forecasts end to end on an embedded device without per-signal fitting.
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。