Sample Efficiency
A metric measuring the amount of interaction data a machine learning model requires to reach a target level of performance. Higher sample efficiency means achieving equivalent results with less data. It is a key bottleneck for deploying reinforcement learning in real-world scenarios.
Timeline (last 90 days)
论文通过控制实验（同一MLP架构、相同超参数搜索空间和训练轮数，仅输入表示不同）证明，Assur导出的特征在样本效率上具有几何平均1.8倍的优势，覆盖50到10000个机构的训练预算。
样本效率指模型达到某一性能水平所需的训练样本数量；几何平均1.8倍优势意味着使用Assur特征的模型达到相同预测误差所需样本数约为基线的1/1.8。
A single farm has only about 18 months of 5-minute-resolution data, and probabilistic model calibration requires more data than deterministic model fitting, so a physics simulator is relied upon to generate synthetic data and fine-tuning is done with a domain adaptation loss.
Sample efficiency and inference efficiency are two independent metrics; CoPE saves interactions during training and maintains single-step actions during inference, with no additional overhead in either.
All Facts (4)
论文通过控制实验（同一MLP架构、相同超参数搜索空间和训练轮数，仅输入表示不同）证明，Assur导出的特征在样本效率上具有几何平均1.8倍的优势，覆盖50到10000个机构的训练预算。
50%Unverified样本效率指模型达到某一性能水平所需的训练样本数量；几何平均1.8倍优势意味着使用Assur特征的模型达到相同预测误差所需样本数约为基线的1/1.8。
50%UnverifiedA single farm has only about 18 months of 5-minute-resolution data, and probabilistic model calibration requires more data than deterministic model fitting, so a physics simulator is relied upon to generate synthetic data and fine-tuning is done with a domain adaptation loss.
50%UnverifiedSample efficiency and inference efficiency are two independent metrics; CoPE saves interactions during training and maintains single-step actions during inference, with no additional overhead in either.
50%