Tag
提出了运动等变卷积 LSTM(MELSTM),将序列数据中的运动视为规范对称性(gauge symmetry),通过速度模型作为规范场实现对任意运动的可证明等变性,在未来帧预测、运动数字分类和动作识别任务上以更少状态超越 ConvLSTM 与流量等变基线。
Benchmarks ConvLSTM against simpler models for one-day-ahead rainfall-field prediction across four Indian cities using IMDAA reanalysis data, finding that ConvLSTM does not consistently outperform alternatives like FC-LSTM or persistence.
This paper presents a GPU-accelerated deep learning framework using ConvLSTM models to predict next-day urban thermal conditions and generate heat risk maps, achieving strong performance (R²=0.8877) on Sarajevo data.