滚动原点预测不稳定性指的是当新数据点可用时更新预测,导致特定时期的预测结果存在变异性。近期,针对单变量时间序列点预测的N-BEATS模型被提出扩展版本,将预测稳定性作为除准确率之外的额外优化目标。研究表明,通过最小化同时包含预测误差和预测不稳定性分量的复合损失函数,并使用静态超参数控制稳定性的影响,可以在不损害准确率的前提下获得更稳定的预测结果。本文通过实证研究,探究在训练过程中调整损失权重的动态损失加权算法,能否在不牺牲准确率的前提下进一步提升稳定性。我们证明现有动态损失加权方法可以实现这一目标,并分析了其背后的潜在原因。此外,我们提出了随机加权(Random Weighting)方法的扩展版本——任务感知随机加权(Task-Aware Random Weighting),该方法同样能够实现上述目标。
英文摘要:
Rolling origin forecast instability refers to variability in forecasts for a specific period induced by updating the forecast when new data points become available. Recently, an extension to the N-BEATS model for univariate time series point forecasting was proposed to include forecast stability as an additional optimization objective, next to accuracy. It was shown that more stable forecasts can be obtained without harming accuracy by minimizing a composite loss function that contains both a forecast error and a forecast instability component, with a static hyperparameter to control the impact of stability. In this paper, we empirically investigate whether further improvements in stability can be obtained without compromising accuracy by applying dynamic loss weighting algorithms, which change the loss weights during training. We show that existing dynamic loss weighting methods can achieve this objective and provide insights into why this might be the case. Additionally, we propose an extension to the Random Weighting approach -- Task-Aware Random Weighting -- which also achieves this objective.
QGAPHEnsemble : Combining Hybrid QLSTM Network Ensemble via Adaptive Weighting for Short Term Weather Forecasting
Georgia Institute of Technology(佐治亚理工学院);Indian Institute of Engineering Science and Technology, Shibpur(印度工程科学技术学院(希布普尔));IIT Roorkee(鲁基印度理工学院);National Institute of Technology, Rourkela(罗尔克拉国家理工学院);National Institute of Technology Karnataka(卡纳塔克邦国家理工学院);Jadavpur University(贾达沃普尔大学)
Quadratic Direct Forecast for Training Multi-Step Time-Series Forecast Models
Xiaohongshu Inc.(小红书公司);State Key Lab of General AI, School of Intelligence Science and Technology, Peking University(人工智能国家重点实验室,北京大学智能科学与技术学院);College of Engineering, Purdue University(普渡大学工程学院);School of Computing and Artificial Intelligence, Shanghai University of Finance and Economics(上海财经大学计算机与人工智能学院);College of Computer Science and Technology, Zhejiang University(浙江大学计算机科学与技术学院);Squirrel AI;Center for Data Science, Peking University(北京大学数据科学中心);Institute for Artificial Intelligence, Peking University(北京大学人工智能研究院);Pazhou Laboratory (Huangpu), Guangzhou, Guangdong, China(琶洲实验室(黄埔),广州,广东,中国)
本文提出二次加权训练目标,解决多步时间序列预测中标签自相关和任务权重不均的问题,通过Quadratic Direct Forecast算法提升模型性能。
ExtremeCast: Boosting Extreme Value Prediction for Global Weather Forecast
Shanghai Artificial Intelligence Laboratory(上海人工智能实验室);Shanghai Jiao Tong University(上海交通大学);University of Science and Technology of China(中国科学技术大学);The Hong Kong University of Science and Technology(香港科技大学)
Adaptive Rainfall Forecasting from Multiple Geographical Models Using Matrix Profile and Ensemble Learning
Ha Noi University of Science and Technology(河内科学技术大学);Center of Environmental Intelligence(环境智能中心);VinUniversity(文大学);University of Southern California(南加州大学)
Weighting What Matters: Boosting Sample Efficiency in Medical Report Generation via Token Reweighting
TUM School of Computation, Information and Technology(慕尼黑技术大学计算、信息与技术学院);Technical University of Munich(慕尼黑技术大学);Munich Center for Machine Learning (MCML)(慕尼黑机器学习中心);Department of Computing, Imperial College London(伦敦帝国学院计算机系)
Efficient Generative AI Boosts Probabilistic Forecasting of Sudden Stratospheric Warmings
Beijing Normal University(北京师范大学);Institute of Atmospheric Physics, Chinese Academy of Sciences(中国科学院大气物理研究所);Zhejiang University(浙江大学);Chinese Academy of Meteorological Sciences(中国气象科学研究院);Lanzhou University(兰州大学)