Research map / Time series / Long-term forecasting
Tensorized LSTM with Adaptive Shared Memory for Learning Trends in Multivariate Time Series
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace basic LSTM with tensorized LSTM + adaptive shared memory (TLASM) (ablation) | DeepTrends · Traffic mu=10 duration | RMSE: 2.2076 -> 1.7587 | Helped |
| replace basic LSTM with TLASM (ablation) | DeepTrends · Solar-Power mu=800 duration | RMSE: 20.7890 -> 13.5987 | Helped |
| replace basic LSTM with TLASM (ablation) | DeepTrends · Traffic mu=10 slope | RMSE: 0.0788 -> 0.0763 | Helped |
| use multi-task 1dCNN instead of basic 1dCNN (ablation) | DeepTrends · Solar-Power mu=800 duration | RMSE: 15.6963 -> 13.5987 | Helped |
| use multi-task 1dCNN instead of basic 1dCNN (ablation) | DeepTrends · Traffic mu=10 duration | RMSE: 1.8120 -> 1.7587 | Helped |
| use multi-task 1dCNN instead of basic 1dCNN (ablation) | DeepTrends · Exchange-Rate mu=6 duration | RMSE: 5.8344 -> 5.1705 | Helped |
| adaptive shared memory vs enhanced shared memory (ME-LSTM) | DeepTrends · Solar-Power mu=800 duration | RMSE: 19.4335 -> 13.5987 | Helped |
| multi-task LSTM-m instead of per-series LSTM | LSTM · Traffic mu=50 duration | RMSE: 3.2263 -> 3.2469 | Hurt |
| multi-task LSTM-m instead of per-series LSTM | LSTM · Solar-Power mu=400 duration | RMSE: 8.9872 -> 8.0060 | Helped |
| DeepTrends vs TreNet (univariate long+short) | TreNet · Traffic mu=10 slope | RMSE: 0.0888 -> 0.0763 | Helped |
| increase window size T and local data size w | DeepTrends · Traffic/Exchange/Solar duration RMSE | RMSE | Mixed |
| add non-zero penalization weight lambda (but large lambda) | DeepTrends · duration RMSE | RMSE | Mixed |
Closest papers by meaning
- TStruct: Learning Shared Temporal Structures for Long-Term Time-Series Forecasting (NeurIPS 2026 · similarity 0.87)
- Unlocking the Power of LSTM for Long Term Time Series Forecasting (AAAI 2025 · similarity 0.86)
- TimeCMA: Towards LLM-Empowered Multivariate Time Series Forecasting via Cross-Modality Alignment (AAAI 2025 · similarity 0.86)
- Time-LLM: Time Series Forecasting by Reprogramming Large Language Models (ICLR 2024 · similarity 0.85)
- T-Rep: Representation Learning for Time Series using Time-Embeddings (ICLR 2024 · similarity 0.85)
- FSTLLM: Spatio-Temporal LLM for Few Shot Time Series Forecasting (ICML 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.