CNN-Based Multi-In-Multi-Out Model for Efficient Spatiotemporal Prediction
A new CNN-based model called Multi-In-Multi-Out (MIMO) is proposed to address limitations in spatiotemporal prediction. Existing CNN models struggle with global information due to local kernel properties and mix time and channel axes, while Transformer models suffer from high complexity and long training times due to self-attention. The MIMO model aims to overcome these challenges by introducing a novel structure that improves efficiency and performance. The paper is published on arXiv with identifier 2605.01277.
Key facts
- The model is called CNN-based Multi-In-Multi-Out (MIMO).
- It targets spatiotemporal prediction tasks.
- CNN models have difficulty with global information due to local kernel properties.
- Transformer models have high complexity from self-attention calculations.
- The paper is available on arXiv under ID 2605.01277.
- The model aims to improve efficiency and performance over existing approaches.
Entities
Institutions
- arXiv