Multi-modal Fusion based Q-distribution Prediction for Controlled Nuclear Fusion

TL;DR

This paper explores multimodal fusion methods in computer vision, integrating 2D line image data with the original 1D data to form a bimodal input and employs the Transformer's attention mechanism for feature extraction and the interactive fusion of bimodal information.

Abstract

Q-distribution prediction is a crucial research direction in controlled nuclear fusion, with deep learning emerging as a key approach to solving prediction challenges. In this paper, we leverage deep learning techniques to tackle the complexities of Q-distribution prediction. Specifically, we explore multimodal fusion methods in computer vision, integrating 2D line image data with the original 1D data to form a bimodal input. Additionally, we employ the Transformer's attention mechanism for feature extraction and the interactive fusion of bimodal information. Extensive experiments validate the effectiveness of our approach, significantly reducing prediction errors in Q-distribution.

Figures (3)

• Figure 1: An overview of our proposed model for the Q-distribution.
• Figure 2: Visual comparison of different resolutions of input 2D images.
• Figure 3: Visual comparison of the number of different multimodal Transformer blocks.